1
0
mirror of https://github.com/moparisthebest/pacman synced 2024-11-11 03:54:59 -05:00
pacman/lib/libalpm/be_local.c

1030 lines
26 KiB
C
Raw Normal View History

/*
* be_local.c : backend for the local database
*
* Copyright (c) 2006-2012 Pacman Development Team <pacman-dev@archlinux.org>
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
* Copyright (c) 2002-2006 by Judd Vinet <jvinet@zeroflux.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include <stdint.h> /* intmax_t */
#include <sys/stat.h>
#include <dirent.h>
#include <limits.h> /* PATH_MAX */
/* libalpm */
#include "db.h"
#include "alpm_list.h"
#include "log.h"
#include "util.h"
#include "alpm.h"
#include "handle.h"
#include "package.h"
#include "deps.h"
#include "filelist.h"
static int local_db_read(alpm_pkg_t *info, alpm_dbinfrq_t inforeq);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
#define LAZY_LOAD(info, errret) \
do { \
if(!(pkg->infolevel & info)) { \
local_db_read(pkg, info); \
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
} \
} while(0)
/* Cache-specific accessor functions. These implementations allow for lazy
* loading by the files backend when a data member is actually needed
* rather than loading all pieces of information when the package is first
* initialized.
*/
static const char *_cache_get_desc(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->desc;
}
static const char *_cache_get_url(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->url;
}
static alpm_time_t _cache_get_builddate(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, 0);
return pkg->builddate;
}
static alpm_time_t _cache_get_installdate(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, 0);
return pkg->installdate;
}
static const char *_cache_get_packager(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->packager;
}
static const char *_cache_get_arch(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->arch;
}
static off_t _cache_get_isize(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, -1);
return pkg->isize;
}
static alpm_pkgreason_t _cache_get_reason(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, -1);
return pkg->reason;
}
static alpm_pkgvalidation_t _cache_get_validation(alpm_pkg_t *pkg)
{
LAZY_LOAD(INFRQ_DESC, -1);
return pkg->validation;
}
static alpm_list_t *_cache_get_licenses(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->licenses;
}
static alpm_list_t *_cache_get_groups(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->groups;
}
static int _cache_has_scriptlet(alpm_pkg_t *pkg)
{
LAZY_LOAD(INFRQ_SCRIPTLET, NULL);
return pkg->scriptlet;
}
static alpm_list_t *_cache_get_depends(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
return pkg->depends;
}
static alpm_list_t *_cache_get_optdepends(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
return pkg->optdepends;
}
static alpm_list_t *_cache_get_conflicts(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
return pkg->conflicts;
}
static alpm_list_t *_cache_get_provides(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
return pkg->provides;
}
static alpm_list_t *_cache_get_replaces(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_DESC, NULL);
return pkg->replaces;
}
static alpm_filelist_t *_cache_get_files(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_FILES, NULL);
return &(pkg->files);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
}
static alpm_list_t *_cache_get_backup(alpm_pkg_t *pkg)
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
{
LAZY_LOAD(INFRQ_FILES, NULL);
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
return pkg->backup;
}
/**
* Open a package changelog for reading. Similar to fopen in functionality,
* except that the returned 'file stream' is from the database.
* @param pkg the package (from db) to read the changelog
* @return a 'file stream' to the package changelog
*/
static void *_cache_changelog_open(alpm_pkg_t *pkg)
{
alpm_db_t *db = alpm_pkg_get_db(pkg);
char *clfile = _alpm_local_db_pkgpath(db, pkg, "changelog");
FILE *f = fopen(clfile, "r");
free(clfile);
return f;
}
/**
* Read data from an open changelog 'file stream'. Similar to fread in
* functionality, this function takes a buffer and amount of data to read.
* @param ptr a buffer to fill with raw changelog data
* @param size the size of the buffer
* @param pkg the package that the changelog is being read from
* @param fp a 'file stream' to the package changelog
* @return the number of characters read, or 0 if there is no more data
*/
static size_t _cache_changelog_read(void *ptr, size_t size,
const alpm_pkg_t UNUSED *pkg, void *fp)
{
return fread(ptr, 1, size, (FILE *)fp);
}
/**
* Close a package changelog for reading. Similar to fclose in functionality,
* except that the 'file stream' is from the database.
* @param pkg the package that the changelog was read from
* @param fp a 'file stream' to the package changelog
* @return whether closing the package changelog stream was successful
*/
static int _cache_changelog_close(const alpm_pkg_t UNUSED *pkg, void *fp)
{
return fclose((FILE *)fp);
}
static int _cache_force_load(alpm_pkg_t *pkg)
{
return local_db_read(pkg, INFRQ_ALL);
}
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
/** The local database operations struct. Get package fields through
* lazy accessor methods that handle any backend loading and caching
* logic.
*/
static struct pkg_operations local_pkg_ops = {
.get_desc = _cache_get_desc,
.get_url = _cache_get_url,
.get_builddate = _cache_get_builddate,
.get_installdate = _cache_get_installdate,
.get_packager = _cache_get_packager,
.get_arch = _cache_get_arch,
.get_isize = _cache_get_isize,
.get_reason = _cache_get_reason,
.get_validation = _cache_get_validation,
.has_scriptlet = _cache_has_scriptlet,
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
.get_licenses = _cache_get_licenses,
.get_groups = _cache_get_groups,
.get_depends = _cache_get_depends,
.get_optdepends = _cache_get_optdepends,
.get_conflicts = _cache_get_conflicts,
.get_provides = _cache_get_provides,
.get_replaces = _cache_get_replaces,
.get_files = _cache_get_files,
.get_backup = _cache_get_backup,
.changelog_open = _cache_changelog_open,
.changelog_read = _cache_changelog_read,
.changelog_close = _cache_changelog_close,
.force_load = _cache_force_load,
Complete rework of package accessor logic Hopefully we've finally arrived at package handling nirvana, or at least this commit will get us a heck of a lot closer. The former method of getting the depends list for a package was the following: 1. call alpm_pkg_get_depends() 2. this method would check if the package came from the cache 3. if so, ensure our cache level is correct, otherwise call db_load 4. finally return the depends list Why did this suck? Because getting the depends list from the package shouldn't care about whether the package was loaded from a file, from the 'package cache', or some other system which we can't even use because the damn thing is so complicated. It should just return the depends list. So what does this commit change? It adds a pointer to a struct of function pointers to every package for all of these 'package operations' as I've decided to call them (I know, sounds completely straightforward, right?). So now when we call an alpm_pkg_get-* function, we don't do any of the cache logic or anything else there- we let the actual backend handle it by delegating all work to the method at pkg->ops->get_depends. Now that be_package has achieved equal status with be_files, we can treat packages from these completely different load points differently. We know a package loaded from a zip file will have all of its fields populated, so we can set up all its accessor functions to be direct accessors. On the other hand, the packages loaded from the local and sync DBs are not always fully-loaded, so their accessor functions are routed through the same logic as before. Net result? More code. However, this code now make it roughly 52 times easier to open the door to something like a read-only tar.gz database backend. Are you still reading? I'm impressed. Looking at the patch will probably be clearer than this long-winded explanation. Signed-off-by: Dan McGee <dan@archlinux.org> [Allan: rebase and adjust] Signed-off-by: Allan McRae <allan@archlinux.org>
2008-05-11 17:00:33 -04:00
};
static int checkdbdir(alpm_db_t *db)
{
struct stat buf;
const char *path = _alpm_db_path(db);
if(stat(path, &buf) != 0) {
_alpm_log(db->handle, ALPM_LOG_DEBUG, "database dir '%s' does not exist, creating it\n",
path);
if(_alpm_makepath(path) != 0) {
RET_ERR(db->handle, ALPM_ERR_SYSTEM, -1);
}
} else if(!S_ISDIR(buf.st_mode)) {
_alpm_log(db->handle, ALPM_LOG_WARNING, _("removing invalid database: %s\n"), path);
if(unlink(path) != 0 || _alpm_makepath(path) != 0) {
RET_ERR(db->handle, ALPM_ERR_SYSTEM, -1);
}
}
return 0;
}
static int is_dir(const char *path, struct dirent *entry)
{
#ifdef HAVE_STRUCT_DIRENT_D_TYPE
if(entry->d_type != DT_UNKNOWN) {
return (entry->d_type == DT_DIR);
}
#endif
{
char buffer[PATH_MAX];
struct stat sbuf;
snprintf(buffer, PATH_MAX, "%s/%s", path, entry->d_name);
if(!stat(buffer, &sbuf)) {
return S_ISDIR(sbuf.st_mode);
}
}
return 0;
}
static int local_db_validate(alpm_db_t *db)
{
struct dirent *ent = NULL;
const char *dbpath;
DIR *dbdir;
int ret = -1;
if(db->status & DB_STATUS_VALID) {
return 0;
}
if(db->status & DB_STATUS_INVALID) {
return -1;
}
dbpath = _alpm_db_path(db);
if(dbpath == NULL) {
RET_ERR(db->handle, ALPM_ERR_DB_OPEN, -1);
}
dbdir = opendir(dbpath);
if(dbdir == NULL) {
if(errno == ENOENT) {
/* database dir doesn't exist yet */
db->status |= DB_STATUS_VALID;
db->status &= ~DB_STATUS_INVALID;
db->status &= ~DB_STATUS_EXISTS;
db->status |= DB_STATUS_MISSING;
return 0;
} else {
RET_ERR(db->handle, ALPM_ERR_DB_OPEN, -1);
}
}
db->status |= DB_STATUS_EXISTS;
db->status &= ~DB_STATUS_MISSING;
while((ent = readdir(dbdir)) != NULL) {
const char *name = ent->d_name;
char path[PATH_MAX];
if(strcmp(name, ".") == 0 || strcmp(name, "..") == 0) {
continue;
}
if(!is_dir(dbpath, ent)) {
continue;
}
snprintf(path, PATH_MAX, "%s%s/depends", dbpath, name);
if(access(path, F_OK) == 0) {
/* we found a depends file- bail */
db->status &= ~DB_STATUS_VALID;
db->status |= DB_STATUS_INVALID;
db->handle->pm_errno = ALPM_ERR_DB_VERSION;
goto done;
}
}
/* we found no depends file after full scan */
db->status |= DB_STATUS_VALID;
db->status &= ~DB_STATUS_INVALID;
ret = 0;
done:
if(dbdir) {
closedir(dbdir);
}
return ret;
}
static int local_db_populate(alpm_db_t *db)
{
size_t est_count;
int count = 0;
struct stat buf;
struct dirent *ent = NULL;
const char *dbpath;
DIR *dbdir;
if(db->status & DB_STATUS_INVALID) {
RET_ERR(db->handle, ALPM_ERR_DB_INVALID, -1);
}
/* note: DB_STATUS_MISSING is not fatal for local database */
dbpath = _alpm_db_path(db);
if(dbpath == NULL) {
/* pm_errno set in _alpm_db_path() */
return -1;
}
dbdir = opendir(dbpath);
if(dbdir == NULL) {
if(errno == ENOENT) {
/* no database existing yet is not an error */
db->status &= ~DB_STATUS_EXISTS;
db->status |= DB_STATUS_MISSING;
return 0;
}
RET_ERR(db->handle, ALPM_ERR_DB_OPEN, -1);
}
if(fstat(dirfd(dbdir), &buf) != 0) {
RET_ERR(db->handle, ALPM_ERR_DB_OPEN, -1);
}
db->status |= DB_STATUS_EXISTS;
db->status &= ~DB_STATUS_MISSING;
if(buf.st_nlink >= 2) {
est_count = buf.st_nlink;
} else {
/* Some filesystems don't subscribe to the two-implicit links school of
* thought, e.g. BTRFS, HFS+. See
* http://kerneltrap.org/mailarchive/linux-btrfs/2010/1/23/6723483/thread
*/
est_count = 0;
while(readdir(dbdir) != NULL) {
est_count++;
}
rewinddir(dbdir);
}
if(est_count >= 2) {
/* subtract the '.' and '..' pointers to get # of children */
est_count -= 2;
}
db->pkgcache = _alpm_pkghash_create(est_count);
if(db->pkgcache == NULL){
closedir(dbdir);
RET_ERR(db->handle, ALPM_ERR_MEMORY, -1);
}
while((ent = readdir(dbdir)) != NULL) {
const char *name = ent->d_name;
alpm_pkg_t *pkg;
if(strcmp(name, ".") == 0 || strcmp(name, "..") == 0) {
continue;
}
if(!is_dir(dbpath, ent)) {
continue;
}
pkg = _alpm_pkg_new();
if(pkg == NULL) {
closedir(dbdir);
RET_ERR(db->handle, ALPM_ERR_MEMORY, -1);
}
/* split the db entry name */
if(_alpm_splitname(name, &(pkg->name), &(pkg->version),
&(pkg->name_hash)) != 0) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("invalid name for database entry '%s'\n"),
name);
_alpm_pkg_free(pkg);
continue;
}
/* duplicated database entries are not allowed */
if(_alpm_pkghash_find(db->pkgcache, pkg->name)) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("duplicated database entry '%s'\n"), pkg->name);
_alpm_pkg_free(pkg);
continue;
}
pkg->origin = ALPM_PKG_FROM_LOCALDB;
pkg->origin_data.db = db;
pkg->ops = &local_pkg_ops;
pkg->handle = db->handle;
/* explicitly read with only 'BASE' data, accessors will handle the rest */
if(local_db_read(pkg, INFRQ_BASE) == -1) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("corrupted database entry '%s'\n"), name);
_alpm_pkg_free(pkg);
continue;
}
/* add to the collection */
_alpm_log(db->handle, ALPM_LOG_FUNCTION, "adding '%s' to package cache for db '%s'\n",
pkg->name, db->treename);
db->pkgcache = _alpm_pkghash_add(db->pkgcache, pkg);
count++;
}
closedir(dbdir);
if(count > 0) {
db->pkgcache->list = alpm_list_msort(db->pkgcache->list, (size_t)count, _alpm_pkg_cmp);
}
_alpm_log(db->handle, ALPM_LOG_DEBUG, "added %d packages to package cache for db '%s'\n",
count, db->treename);
return count;
}
/* Note: the return value must be freed by the caller */
char *_alpm_local_db_pkgpath(alpm_db_t *db, alpm_pkg_t *info,
const char *filename)
{
size_t len;
char *pkgpath;
const char *dbpath;
dbpath = _alpm_db_path(db);
len = strlen(dbpath) + strlen(info->name) + strlen(info->version) + 3;
len += filename ? strlen(filename) : 0;
MALLOC(pkgpath, len, RET_ERR(db->handle, ALPM_ERR_MEMORY, NULL));
sprintf(pkgpath, "%s%s-%s/%s", dbpath, info->name, info->version,
filename ? filename : "");
return pkgpath;
}
#define READ_NEXT() do { \
if(fgets(line, sizeof(line), fp) == NULL && !feof(fp)) goto error; \
_alpm_strip_newline(line, 0); \
} while(0)
#define READ_AND_STORE(f) do { \
READ_NEXT(); \
STRDUP(f, line, goto error); \
} while(0)
#define READ_AND_STORE_ALL(f) do { \
char *linedup; \
if(fgets(line, sizeof(line), fp) == NULL) {\
if(!feof(fp)) goto error; else break; \
} \
if(_alpm_strip_newline(line, 0) == 0) break; \
STRDUP(linedup, line, goto error); \
f = alpm_list_add(f, linedup); \
} while(1) /* note the while(1) and not (0) */
#define READ_AND_SPLITDEP(f) do { \
if(fgets(line, sizeof(line), fp) == NULL) {\
if(!feof(fp)) goto error; else break; \
} \
if(_alpm_strip_newline(line, 0) == 0) break; \
f = alpm_list_add(f, _alpm_splitdep(line)); \
} while(1) /* note the while(1) and not (0) */
static int local_db_read(alpm_pkg_t *info, alpm_dbinfrq_t inforeq)
{
FILE *fp = NULL;
char line[1024];
alpm_db_t *db = info->origin_data.db;
/* bitmask logic here:
* infolevel: 00001111
* inforeq: 00010100
* & result: 00000100
* == to inforeq? nope, we need to load more info. */
if((info->infolevel & inforeq) == inforeq) {
/* already loaded all of this info, do nothing */
return 0;
}
if(info->infolevel & INFRQ_ERROR) {
/* We've encountered an error loading this package before. Don't attempt
* repeated reloads, just give up. */
return -1;
}
_alpm_log(db->handle, ALPM_LOG_FUNCTION,
"loading package data for %s : level=0x%x\n",
info->name, inforeq);
/* clear out 'line', to be certain - and to make valgrind happy */
memset(line, 0, sizeof(line));
/* DESC */
if(inforeq & INFRQ_DESC && !(info->infolevel & INFRQ_DESC)) {
char *path = _alpm_local_db_pkgpath(db, info, "desc");
if(!path || (fp = fopen(path, "r")) == NULL) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("could not open file %s: %s\n"), path, strerror(errno));
free(path);
goto error;
}
free(path);
while(!feof(fp)) {
if(fgets(line, sizeof(line), fp) == NULL && !feof(fp)) {
goto error;
}
if(_alpm_strip_newline(line, 0) == 0) {
/* length of stripped line was zero */
continue;
}
if(strcmp(line, "%NAME%") == 0) {
READ_NEXT();
if(strcmp(line, info->name) != 0) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("%s database is inconsistent: name "
"mismatch on package %s\n"), db->treename, info->name);
}
} else if(strcmp(line, "%VERSION%") == 0) {
READ_NEXT();
if(strcmp(line, info->version) != 0) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("%s database is inconsistent: version "
"mismatch on package %s\n"), db->treename, info->name);
}
} else if(strcmp(line, "%DESC%") == 0) {
READ_AND_STORE(info->desc);
} else if(strcmp(line, "%GROUPS%") == 0) {
READ_AND_STORE_ALL(info->groups);
} else if(strcmp(line, "%URL%") == 0) {
READ_AND_STORE(info->url);
} else if(strcmp(line, "%LICENSE%") == 0) {
READ_AND_STORE_ALL(info->licenses);
} else if(strcmp(line, "%ARCH%") == 0) {
READ_AND_STORE(info->arch);
} else if(strcmp(line, "%BUILDDATE%") == 0) {
READ_NEXT();
info->builddate = _alpm_parsedate(line);
} else if(strcmp(line, "%INSTALLDATE%") == 0) {
READ_NEXT();
info->installdate = _alpm_parsedate(line);
} else if(strcmp(line, "%PACKAGER%") == 0) {
READ_AND_STORE(info->packager);
} else if(strcmp(line, "%REASON%") == 0) {
READ_NEXT();
info->reason = (alpm_pkgreason_t)atoi(line);
} else if(strcmp(line, "%VALIDATION%") == 0) {
alpm_list_t *i, *v = NULL;
READ_AND_STORE_ALL(v);
for(i = v; i; i = alpm_list_next(i))
{
if(strcmp(i->data, "none") == 0) {
info->validation |= ALPM_PKG_VALIDATION_NONE;
} else if(strcmp(i->data, "md5") == 0) {
info->validation |= ALPM_PKG_VALIDATION_MD5SUM;
} else if(strcmp(i->data, "sha256") == 0) {
info->validation |= ALPM_PKG_VALIDATION_SHA256SUM;
} else if(strcmp(i->data, "pgp") == 0) {
info->validation |= ALPM_PKG_VALIDATION_SIGNATURE;
} else {
_alpm_log(db->handle, ALPM_LOG_WARNING,
_("unknown validation type for package %s: %s\n"),
info->name, (const char *)i->data);
}
}
FREELIST(v);
} else if(strcmp(line, "%SIZE%") == 0) {
READ_NEXT();
info->isize = _alpm_strtoofft(line);
} else if(strcmp(line, "%REPLACES%") == 0) {
READ_AND_SPLITDEP(info->replaces);
} else if(strcmp(line, "%DEPENDS%") == 0) {
READ_AND_SPLITDEP(info->depends);
} else if(strcmp(line, "%OPTDEPENDS%") == 0) {
READ_AND_SPLITDEP(info->optdepends);
} else if(strcmp(line, "%CONFLICTS%") == 0) {
READ_AND_SPLITDEP(info->conflicts);
} else if(strcmp(line, "%PROVIDES%") == 0) {
READ_AND_SPLITDEP(info->provides);
}
}
fclose(fp);
fp = NULL;
info->infolevel |= INFRQ_DESC;
}
/* FILES */
if(inforeq & INFRQ_FILES && !(info->infolevel & INFRQ_FILES)) {
char *path = _alpm_local_db_pkgpath(db, info, "files");
if(!path || (fp = fopen(path, "r")) == NULL) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("could not open file %s: %s\n"), path, strerror(errno));
free(path);
goto error;
}
free(path);
while(fgets(line, sizeof(line), fp)) {
_alpm_strip_newline(line, 0);
if(strcmp(line, "%FILES%") == 0) {
size_t files_count = 0, files_size = 0, len;
alpm_file_t *files = NULL;
while(fgets(line, sizeof(line), fp) &&
(len = _alpm_strip_newline(line, 0))) {
if(files_count >= files_size) {
size_t old_size = files_size;
if(files_size == 0) {
files_size = 8;
} else {
files_size *= 2;
}
files = realloc(files, sizeof(alpm_file_t) * files_size);
if(!files) {
_alpm_alloc_fail(sizeof(alpm_file_t) * files_size);
goto error;
}
/* ensure all new memory is zeroed out, in both the initial
* allocation and later reallocs */
memset(files + old_size, 0,
sizeof(alpm_file_t) * (files_size - old_size));
}
/* since we know the length of the file string already,
* we can do malloc + memcpy rather than strdup */
len += 1;
files[files_count].name = malloc(len);
if(files[files_count].name == NULL) {
_alpm_alloc_fail(len);
goto error;
}
memcpy(files[files_count].name, line, len);
files_count++;
}
/* attempt to hand back any memory we don't need */
files = realloc(files, sizeof(alpm_file_t) * files_count);
/* make sure the list is sorted */
qsort(files, files_count, sizeof(alpm_file_t), _alpm_files_cmp);
info->files.count = files_count;
info->files.files = files;
} else if(strcmp(line, "%BACKUP%") == 0) {
while(fgets(line, sizeof(line), fp) && _alpm_strip_newline(line, 0)) {
alpm_backup_t *backup;
CALLOC(backup, 1, sizeof(alpm_backup_t), goto error);
if(_alpm_split_backup(line, &backup)) {
goto error;
}
info->backup = alpm_list_add(info->backup, backup);
}
}
}
fclose(fp);
fp = NULL;
info->infolevel |= INFRQ_FILES;
}
/* INSTALL */
if(inforeq & INFRQ_SCRIPTLET && !(info->infolevel & INFRQ_SCRIPTLET)) {
char *path = _alpm_local_db_pkgpath(db, info, "install");
if(access(path, F_OK) == 0) {
info->scriptlet = 1;
}
free(path);
info->infolevel |= INFRQ_SCRIPTLET;
}
return 0;
error:
info->infolevel |= INFRQ_ERROR;
if(fp) {
fclose(fp);
}
return -1;
}
int _alpm_local_db_prepare(alpm_db_t *db, alpm_pkg_t *info)
{
mode_t oldmask;
int retval = 0;
char *pkgpath;
if(checkdbdir(db) != 0) {
return -1;
}
oldmask = umask(0000);
pkgpath = _alpm_local_db_pkgpath(db, info, NULL);
if((retval = mkdir(pkgpath, 0755)) != 0) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("could not create directory %s: %s\n"),
pkgpath, strerror(errno));
}
free(pkgpath);
umask(oldmask);
return retval;
}
static void write_deps(FILE *fp, const char *header, alpm_list_t *deplist)
{
alpm_list_t *lp;
if(!deplist) {
return;
}
fputs(header, fp);
fputc('\n', fp);
for(lp = deplist; lp; lp = lp->next) {
char *depstring = alpm_dep_compute_string(lp->data);
fputs(depstring, fp);
fputc('\n', fp);
free(depstring);
}
fputc('\n', fp);
}
int _alpm_local_db_write(alpm_db_t *db, alpm_pkg_t *info, alpm_dbinfrq_t inforeq)
{
FILE *fp = NULL;
mode_t oldmask;
alpm_list_t *lp;
int retval = 0;
if(db == NULL || info == NULL || !(db->status & DB_STATUS_LOCAL)) {
return -1;
}
/* make sure we have a sane umask */
oldmask = umask(0022);
/* DESC */
if(inforeq & INFRQ_DESC) {
char *path;
_alpm_log(db->handle, ALPM_LOG_DEBUG,
"writing %s-%s DESC information back to db\n",
info->name, info->version);
path = _alpm_local_db_pkgpath(db, info, "desc");
if(!path || (fp = fopen(path, "w")) == NULL) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("could not open file %s: %s\n"),
path, strerror(errno));
retval = -1;
free(path);
goto cleanup;
}
free(path);
fprintf(fp, "%%NAME%%\n%s\n\n"
"%%VERSION%%\n%s\n\n", info->name, info->version);
if(info->desc) {
fprintf(fp, "%%DESC%%\n"
"%s\n\n", info->desc);
}
if(info->url) {
fprintf(fp, "%%URL%%\n"
"%s\n\n", info->url);
}
if(info->arch) {
fprintf(fp, "%%ARCH%%\n"
"%s\n\n", info->arch);
}
if(info->builddate) {
fprintf(fp, "%%BUILDDATE%%\n"
"%jd\n\n", (intmax_t)info->builddate);
}
if(info->installdate) {
fprintf(fp, "%%INSTALLDATE%%\n"
"%jd\n\n", (intmax_t)info->installdate);
}
if(info->packager) {
fprintf(fp, "%%PACKAGER%%\n"
"%s\n\n", info->packager);
}
if(info->isize) {
/* only write installed size, csize is irrelevant once installed */
fprintf(fp, "%%SIZE%%\n"
"%jd\n\n", (intmax_t)info->isize);
}
if(info->reason) {
fprintf(fp, "%%REASON%%\n"
"%u\n\n", info->reason);
}
if(info->groups) {
fputs("%GROUPS%\n", fp);
for(lp = info->groups; lp; lp = lp->next) {
fputs(lp->data, fp);
fputc('\n', fp);
}
fputc('\n', fp);
}
if(info->licenses) {
fputs("%LICENSE%\n", fp);
for(lp = info->licenses; lp; lp = lp->next) {
fputs(lp->data, fp);
fputc('\n', fp);
}
fputc('\n', fp);
}
if(info->validation) {
fputs("%VALIDATION%\n", fp);
if(info->validation & ALPM_PKG_VALIDATION_NONE) {
fputs("none\n", fp);
}
if(info->validation & ALPM_PKG_VALIDATION_MD5SUM) {
fputs("md5\n", fp);
}
if(info->validation & ALPM_PKG_VALIDATION_SHA256SUM) {
fputs("sha256\n", fp);
}
if(info->validation & ALPM_PKG_VALIDATION_SIGNATURE) {
fputs("pgp\n", fp);
}
fputc('\n', fp);
}
write_deps(fp, "%REPLACES%", info->replaces);
write_deps(fp, "%DEPENDS%", info->depends);
write_deps(fp, "%OPTDEPENDS%", info->optdepends);
write_deps(fp, "%CONFLICTS%", info->conflicts);
write_deps(fp, "%PROVIDES%", info->provides);
fclose(fp);
fp = NULL;
}
/* FILES */
if(inforeq & INFRQ_FILES) {
char *path;
_alpm_log(db->handle, ALPM_LOG_DEBUG,
"writing %s-%s FILES information back to db\n",
info->name, info->version);
path = _alpm_local_db_pkgpath(db, info, "files");
if(!path || (fp = fopen(path, "w")) == NULL) {
_alpm_log(db->handle, ALPM_LOG_ERROR, _("could not open file %s: %s\n"),
path, strerror(errno));
retval = -1;
free(path);
goto cleanup;
}
free(path);
if(info->files.count) {
size_t i;
fputs("%FILES%\n", fp);
for(i = 0; i < info->files.count; i++) {
const alpm_file_t *file = info->files.files + i;
fputs(file->name, fp);
fputc('\n', fp);
}
fputc('\n', fp);
}
if(info->backup) {
fputs("%BACKUP%\n", fp);
for(lp = info->backup; lp; lp = lp->next) {
const alpm_backup_t *backup = lp->data;
fprintf(fp, "%s\t%s\n", backup->name, backup->hash);
}
fputc('\n', fp);
}
fclose(fp);
fp = NULL;
}
/* INSTALL */
/* nothing needed here (script is automatically extracted) */
cleanup:
umask(oldmask);
if(fp) {
fclose(fp);
}
return retval;
}
int _alpm_local_db_remove(alpm_db_t *db, alpm_pkg_t *info)
{
int ret = 0;
DIR *dirp;
struct dirent *dp;
char *pkgpath;
size_t pkgpath_len;
pkgpath = _alpm_local_db_pkgpath(db, info, NULL);
if(!pkgpath) {
return -1;
}
pkgpath_len = strlen(pkgpath);
dirp = opendir(pkgpath);
if(!dirp) {
free(pkgpath);
return -1;
}
/* go through the local DB entry, removing the files within, which we know
* are not nested directories of any kind. */
for(dp = readdir(dirp); dp != NULL; dp = readdir(dirp)) {
if(strcmp(dp->d_name, "..") != 0 && strcmp(dp->d_name, ".") != 0) {
char name[PATH_MAX];
if(pkgpath_len + strlen(dp->d_name) + 2 > PATH_MAX) {
/* file path is too long to remove, hmm. */
ret = -1;
} else {
sprintf(name, "%s/%s", pkgpath, dp->d_name);
if(unlink(name)) {
ret = -1;
}
}
}
}
closedir(dirp);
/* after removing all enclosed files, we can remove the directory itself. */
if(rmdir(pkgpath)) {
ret = -1;
}
free(pkgpath);
return ret;
}
int SYMEXPORT alpm_pkg_set_reason(alpm_pkg_t *pkg, alpm_pkgreason_t reason)
{
ASSERT(pkg != NULL, return -1);
ASSERT(pkg->origin == ALPM_PKG_FROM_LOCALDB,
RET_ERR(pkg->handle, ALPM_ERR_WRONG_ARGS, -1));
ASSERT(pkg->origin_data.db == pkg->handle->db_local,
RET_ERR(pkg->handle, ALPM_ERR_WRONG_ARGS, -1));
_alpm_log(pkg->handle, ALPM_LOG_DEBUG,
"setting install reason %u for %s\n", reason, pkg->name);
if(alpm_pkg_get_reason(pkg) == reason) {
/* we are done */
return 0;
}
/* set reason (in pkgcache) */
pkg->reason = reason;
/* write DESC */
if(_alpm_local_db_write(pkg->handle->db_local, pkg, INFRQ_DESC)) {
RET_ERR(pkg->handle, ALPM_ERR_DB_WRITE, -1);
}
return 0;
}
struct db_operations local_db_ops = {
.validate = local_db_validate,
.populate = local_db_populate,
.unregister = _alpm_db_unregister,
};
alpm_db_t *_alpm_db_register_local(alpm_handle_t *handle)
{
alpm_db_t *db;
_alpm_log(handle, ALPM_LOG_DEBUG, "registering local database\n");
db = _alpm_db_new("local", 1);
if(db == NULL) {
handle->pm_errno = ALPM_ERR_DB_CREATE;
return NULL;
}
db->ops = &local_db_ops;
db->handle = handle;
if(local_db_validate(db)) {
/* pm_errno set in local_db_validate() */
_alpm_db_free(db);
return NULL;
}
handle->db_local = db;
return db;
}
/* vim: set ts=2 sw=2 noet: */