2008-01-22 22:36:11 -05:00
|
|
|
/*
|
|
|
|
* download.c
|
|
|
|
*
|
2011-01-05 23:45:15 -05:00
|
|
|
* Copyright (c) 2006-2011 Pacman Development Team <pacman-dev@archlinux.org>
|
2009-07-01 03:08:33 -04:00
|
|
|
* Copyright (c) 2002-2006 by Judd Vinet <jvinet@zeroflux.org>
|
2008-01-22 22:36:11 -05:00
|
|
|
*
|
|
|
|
* 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 "config.h"
|
|
|
|
|
|
|
|
#include <stdlib.h>
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <errno.h>
|
|
|
|
#include <string.h>
|
|
|
|
#include <unistd.h>
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
#include <sys/time.h>
|
|
|
|
#include <sys/types.h>
|
|
|
|
#include <sys/stat.h>
|
2008-10-31 20:40:55 -04:00
|
|
|
#include <signal.h>
|
2008-04-19 13:36:18 -04:00
|
|
|
|
2010-12-15 19:57:31 -05:00
|
|
|
#ifdef HAVE_LIBCURL
|
|
|
|
#include <curl/curl.h>
|
|
|
|
#endif
|
|
|
|
|
2008-01-22 22:36:11 -05:00
|
|
|
/* libalpm */
|
|
|
|
#include "dload.h"
|
|
|
|
#include "alpm_list.h"
|
|
|
|
#include "alpm.h"
|
|
|
|
#include "log.h"
|
|
|
|
#include "util.h"
|
|
|
|
#include "handle.h"
|
|
|
|
|
2011-03-17 09:01:30 -04:00
|
|
|
#ifdef HAVE_LIBCURL
|
|
|
|
static double prevprogress; /* last download amount */
|
|
|
|
#endif
|
2011-03-17 09:01:30 -04:00
|
|
|
|
2011-03-17 09:01:30 -04:00
|
|
|
static char *get_filename(const char *url)
|
|
|
|
{
|
2008-02-27 16:36:53 -05:00
|
|
|
char *filename = strrchr(url, '/');
|
|
|
|
if(filename != NULL) {
|
|
|
|
filename++;
|
|
|
|
}
|
2011-03-20 20:45:57 -04:00
|
|
|
return filename;
|
2008-02-27 16:36:53 -05:00
|
|
|
}
|
2008-01-22 23:00:12 -05:00
|
|
|
|
2011-02-08 21:05:53 -05:00
|
|
|
#ifdef HAVE_LIBCURL
|
2011-03-25 08:37:02 -04:00
|
|
|
static char *get_fullpath(const char *path, const char *filename,
|
|
|
|
const char *suffix)
|
2011-02-10 11:19:41 -05:00
|
|
|
{
|
2011-03-25 08:37:02 -04:00
|
|
|
char *filepath;
|
|
|
|
/* len = localpath len + filename len + suffix len + null */
|
|
|
|
size_t len = strlen(path) + strlen(filename) + strlen(suffix) + 1;
|
|
|
|
CALLOC(filepath, len, sizeof(char), RET_ERR(PM_ERR_MEMORY, NULL));
|
|
|
|
snprintf(filepath, len, "%s%s%s", path, filename, suffix);
|
2008-01-22 23:00:12 -05:00
|
|
|
|
2011-03-25 08:37:02 -04:00
|
|
|
return filepath;
|
2008-02-27 16:36:53 -05:00
|
|
|
}
|
|
|
|
|
2011-01-14 12:17:39 -05:00
|
|
|
#define check_stop() if(dload_interrupted) { ret = -1; goto cleanup; }
|
|
|
|
enum sighandlers { OLD = 0, NEW = 1 };
|
|
|
|
|
|
|
|
int dload_interrupted;
|
|
|
|
static void inthandler(int signum)
|
|
|
|
{
|
|
|
|
dload_interrupted = 1;
|
|
|
|
}
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
|
2011-03-21 16:54:00 -04:00
|
|
|
static int curl_progress(void *file, double dltotal, double dlnow,
|
2011-02-10 11:19:41 -05:00
|
|
|
double ultotal, double ulnow)
|
|
|
|
{
|
2011-03-21 16:54:00 -04:00
|
|
|
struct fileinfo *dlfile = (struct fileinfo *)file;
|
|
|
|
double current_size, total_size;
|
2011-02-01 15:38:37 -05:00
|
|
|
|
|
|
|
/* unused parameters */
|
|
|
|
(void)ultotal;
|
|
|
|
(void)ulnow;
|
|
|
|
|
2011-03-17 09:01:30 -04:00
|
|
|
/* SIGINT sent, abort by alerting curl */
|
|
|
|
if(dload_interrupted) {
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* none of what follows matters if the front end has no callback */
|
|
|
|
if(handle->dlcb == NULL) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-03-21 16:54:00 -04:00
|
|
|
current_size = dlfile->initial_size + dlnow;
|
|
|
|
total_size = dlfile->initial_size + dltotal;
|
|
|
|
|
|
|
|
if(DOUBLE_EQ(dltotal, 0) || DOUBLE_EQ(prevprogress, total_size)) {
|
2011-03-20 20:45:57 -04:00
|
|
|
return 0;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
2011-03-17 09:01:30 -04:00
|
|
|
/* initialize the progress bar here to avoid displaying it when
|
|
|
|
* a repo is up to date and nothing gets downloaded */
|
|
|
|
if(DOUBLE_EQ(prevprogress, 0)) {
|
2011-03-21 16:54:00 -04:00
|
|
|
handle->dlcb(dlfile->filename, 0, (long)dltotal);
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
2011-03-21 16:54:00 -04:00
|
|
|
handle->dlcb(dlfile->filename, (long)current_size, (long)total_size);
|
2011-02-01 15:38:37 -05:00
|
|
|
|
2011-03-21 16:54:00 -04:00
|
|
|
prevprogress = current_size;
|
2011-03-17 09:01:30 -04:00
|
|
|
|
2011-03-20 20:45:57 -04:00
|
|
|
return 0;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
2011-02-10 11:19:41 -05:00
|
|
|
static int curl_gethost(const char *url, char *buffer)
|
|
|
|
{
|
2011-02-01 15:38:37 -05:00
|
|
|
int hostlen;
|
|
|
|
char *p;
|
|
|
|
|
|
|
|
if(strncmp(url, "file://", 7) == 0) {
|
|
|
|
strcpy(buffer, _("disk"));
|
|
|
|
} else {
|
|
|
|
p = strstr(url, "//");
|
|
|
|
if(!p) {
|
2011-03-20 20:45:57 -04:00
|
|
|
return 1;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
p += 2; /* jump over the found // */
|
|
|
|
hostlen = strcspn(p, "/");
|
|
|
|
if(hostlen > 255) {
|
|
|
|
/* buffer overflow imminent */
|
|
|
|
_alpm_log(PM_LOG_ERROR, _("buffer overflow detected"));
|
2011-03-20 20:45:57 -04:00
|
|
|
return 1;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
snprintf(buffer, hostlen + 1, "%s", p);
|
|
|
|
}
|
|
|
|
|
2011-03-20 20:45:57 -04:00
|
|
|
return 0;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
static int curl_download_internal(const char *url, const char *localpath,
|
2011-02-10 11:19:41 -05:00
|
|
|
int force)
|
|
|
|
{
|
2011-02-01 15:38:37 -05:00
|
|
|
int ret = -1;
|
|
|
|
FILE *localf = NULL;
|
2011-03-25 08:45:54 -04:00
|
|
|
const char *open_mode, *useragent;
|
|
|
|
char *destfile, *tempfile;
|
2011-02-01 15:38:37 -05:00
|
|
|
char hostname[256]; /* RFC1123 states applications should support this length */
|
|
|
|
struct stat st;
|
2011-03-25 08:45:54 -04:00
|
|
|
long httpresp, timecond, remote_time;
|
2011-02-01 15:38:37 -05:00
|
|
|
double remote_size, bytes_dl;
|
|
|
|
struct sigaction sig_pipe[2], sig_int[2];
|
2011-03-21 16:54:00 -04:00
|
|
|
struct fileinfo dlfile;
|
2011-02-01 15:38:37 -05:00
|
|
|
|
2011-03-21 16:54:00 -04:00
|
|
|
dlfile.initial_size = 0.0;
|
|
|
|
dlfile.filename = get_filename(url);
|
|
|
|
if(!dlfile.filename || curl_gethost(url, hostname) != 0) {
|
2011-02-01 15:38:37 -05:00
|
|
|
_alpm_log(PM_LOG_ERROR, _("url '%s' is invalid\n"), url);
|
|
|
|
RET_ERR(PM_ERR_SERVER_BAD_URL, -1);
|
|
|
|
}
|
|
|
|
|
2011-03-25 08:37:02 -04:00
|
|
|
destfile = get_fullpath(localpath, dlfile.filename, "");
|
|
|
|
tempfile = get_fullpath(localpath, dlfile.filename, ".part");
|
|
|
|
if(!destfile || !tempfile) {
|
|
|
|
goto cleanup;
|
|
|
|
}
|
2011-02-01 15:38:37 -05:00
|
|
|
|
|
|
|
/* the curl_easy handle is initialized with the alpm handle, so we only need
|
|
|
|
* to reset the curl handle set parameters for each time it's used. */
|
|
|
|
curl_easy_reset(handle->curl);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_URL, url);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_FAILONERROR, 1L);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_CONNECTTIMEOUT, 10L);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_FILETIME, 1L);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_NOPROGRESS, 0L);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_FOLLOWLOCATION, 1L);
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_PROGRESSFUNCTION, curl_progress);
|
2011-03-21 16:54:00 -04:00
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_PROGRESSDATA, (void*)&dlfile);
|
2011-02-01 15:38:37 -05:00
|
|
|
|
2011-03-17 18:33:34 -04:00
|
|
|
useragent = getenv("HTTP_USER_AGENT");
|
|
|
|
if (useragent != NULL) {
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_USERAGENT, useragent);
|
|
|
|
}
|
|
|
|
|
2011-03-25 08:45:54 -04:00
|
|
|
/* TODO: no assuming here. the calling function should tell us what's kosher */
|
2011-02-01 15:38:37 -05:00
|
|
|
if(!force && stat(destfile, &st) == 0) {
|
|
|
|
/* assume its a sync, so we're starting from scratch. but, only download
|
|
|
|
* our local is out of date. */
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_TIMECONDITION, CURL_TIMECOND_IFMODSINCE);
|
2011-03-25 08:45:54 -04:00
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_TIMEVALUE, (long)st.st_mtime);
|
2011-02-01 15:38:37 -05:00
|
|
|
} else if(stat(tempfile, &st) == 0 && st.st_size > 0) {
|
|
|
|
/* assume its a partial package download. we do not support resuming of
|
|
|
|
* transfers on partially downloaded sync DBs. */
|
2011-03-25 08:45:54 -04:00
|
|
|
open_mode = "ab";
|
2011-02-01 15:38:37 -05:00
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_RESUME_FROM, (long)st.st_size);
|
|
|
|
_alpm_log(PM_LOG_DEBUG, "tempfile found, attempting continuation");
|
2011-03-21 16:54:00 -04:00
|
|
|
dlfile.initial_size = (double)st.st_size;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
2011-03-25 08:45:54 -04:00
|
|
|
localf = fopen(tempfile, open_mode);
|
2011-02-01 15:38:37 -05:00
|
|
|
if(localf == NULL) {
|
2011-03-25 08:45:54 -04:00
|
|
|
goto cleanup;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
curl_easy_setopt(handle->curl, CURLOPT_WRITEDATA, localf);
|
|
|
|
|
|
|
|
/* print proxy info for debug purposes */
|
|
|
|
_alpm_log(PM_LOG_DEBUG, "HTTP_PROXY: %s\n", getenv("HTTP_PROXY"));
|
|
|
|
_alpm_log(PM_LOG_DEBUG, "http_proxy: %s\n", getenv("http_proxy"));
|
|
|
|
_alpm_log(PM_LOG_DEBUG, "FTP_PROXY: %s\n", getenv("FTP_PROXY"));
|
|
|
|
_alpm_log(PM_LOG_DEBUG, "ftp_proxy: %s\n", getenv("ftp_proxy"));
|
|
|
|
|
|
|
|
/* ignore any SIGPIPE signals- these may occur if our FTP socket dies or
|
|
|
|
* something along those lines. Store the old signal handler first. */
|
|
|
|
sig_pipe[NEW].sa_handler = SIG_IGN;
|
|
|
|
sigemptyset(&sig_pipe[NEW].sa_mask);
|
|
|
|
sig_pipe[NEW].sa_flags = 0;
|
|
|
|
sigaction(SIGPIPE, NULL, &sig_pipe[OLD]);
|
|
|
|
sigaction(SIGPIPE, &sig_pipe[NEW], NULL);
|
|
|
|
|
|
|
|
dload_interrupted = 0;
|
|
|
|
sig_int[NEW].sa_handler = &inthandler;
|
|
|
|
sigemptyset(&sig_int[NEW].sa_mask);
|
|
|
|
sig_int[NEW].sa_flags = 0;
|
|
|
|
sigaction(SIGINT, NULL, &sig_int[OLD]);
|
|
|
|
sigaction(SIGINT, &sig_int[NEW], NULL);
|
|
|
|
|
2011-03-17 09:01:30 -04:00
|
|
|
/* Progress 0 - initialize */
|
|
|
|
prevprogress = 0;
|
2011-02-01 15:38:37 -05:00
|
|
|
|
|
|
|
/* perform transfer */
|
|
|
|
handle->curlerr = curl_easy_perform(handle->curl);
|
|
|
|
|
|
|
|
/* retrieve info about the state of the transfer */
|
2011-03-18 22:29:23 -04:00
|
|
|
curl_easy_getinfo(handle->curl, CURLINFO_RESPONSE_CODE, &httpresp);
|
2011-02-01 15:38:37 -05:00
|
|
|
curl_easy_getinfo(handle->curl, CURLINFO_FILETIME, &remote_time);
|
|
|
|
curl_easy_getinfo(handle->curl, CURLINFO_CONTENT_LENGTH_DOWNLOAD, &remote_size);
|
|
|
|
curl_easy_getinfo(handle->curl, CURLINFO_SIZE_DOWNLOAD, &bytes_dl);
|
|
|
|
curl_easy_getinfo(handle->curl, CURLINFO_CONDITION_UNMET, &timecond);
|
|
|
|
|
|
|
|
/* time condition was met and we didn't download anything. we need to
|
|
|
|
* clean up the 0 byte .part file that's left behind. */
|
2011-03-17 22:56:17 -04:00
|
|
|
if(DOUBLE_EQ(bytes_dl, 0) && timecond == 1) {
|
2011-02-01 15:38:37 -05:00
|
|
|
ret = 1;
|
|
|
|
unlink(tempfile);
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(handle->curlerr == CURLE_ABORTED_BY_CALLBACK) {
|
|
|
|
goto cleanup;
|
|
|
|
} else if(handle->curlerr != CURLE_OK) {
|
|
|
|
pm_errno = PM_ERR_LIBCURL;
|
|
|
|
_alpm_log(PM_LOG_ERROR, _("failed retrieving file '%s' from %s : %s\n"),
|
2011-03-21 16:54:00 -04:00
|
|
|
dlfile.filename, hostname, curl_easy_strerror(handle->curlerr));
|
2011-02-01 15:38:37 -05:00
|
|
|
unlink(tempfile);
|
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* remote_size isn't necessarily the full size of the file, just what the
|
|
|
|
* server reported as remaining to download. compare it to what curl reported
|
|
|
|
* as actually being transferred during curl_easy_perform() */
|
2011-03-17 22:56:17 -04:00
|
|
|
if(!DOUBLE_EQ(remote_size, -1) && !DOUBLE_EQ(bytes_dl, -1) &&
|
|
|
|
!DOUBLE_EQ(bytes_dl, remote_size)) {
|
2011-02-01 15:38:37 -05:00
|
|
|
pm_errno = PM_ERR_RETRIEVE;
|
|
|
|
_alpm_log(PM_LOG_ERROR, _("%s appears to be truncated: %jd/%jd bytes\n"),
|
2011-03-21 16:54:00 -04:00
|
|
|
dlfile.filename, (intmax_t)bytes_dl, (intmax_t)remote_size);
|
2011-02-01 15:38:37 -05:00
|
|
|
goto cleanup;
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(localf);
|
|
|
|
localf = NULL;
|
|
|
|
|
|
|
|
/* set the times on the file to the same as that of the remote file */
|
|
|
|
if(remote_time != -1) {
|
|
|
|
struct timeval tv[2];
|
|
|
|
memset(&tv, 0, sizeof(tv));
|
|
|
|
tv[0].tv_sec = tv[1].tv_sec = remote_time;
|
|
|
|
utimes(tempfile, tv);
|
|
|
|
}
|
|
|
|
rename(tempfile, destfile);
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
cleanup:
|
|
|
|
FREE(tempfile);
|
|
|
|
FREE(destfile);
|
|
|
|
if(localf != NULL) {
|
|
|
|
/* if we still had a local file open, we got interrupted. set the mtimes on
|
|
|
|
* the file accordingly. */
|
|
|
|
fflush(localf);
|
|
|
|
if(remote_time != -1) {
|
|
|
|
struct timeval tv[2];
|
|
|
|
memset(&tv, 0, sizeof(tv));
|
|
|
|
tv[0].tv_sec = tv[1].tv_sec = remote_time;
|
|
|
|
futimes(fileno(localf), tv);
|
|
|
|
}
|
|
|
|
fclose(localf);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* restore the old signal handlers */
|
|
|
|
sigaction(SIGINT, &sig_int[OLD], NULL);
|
|
|
|
sigaction(SIGPIPE, &sig_pipe[OLD], NULL);
|
|
|
|
/* if we were interrupted, trip the old handler */
|
|
|
|
if(dload_interrupted) {
|
|
|
|
raise(SIGINT);
|
|
|
|
}
|
|
|
|
|
2011-03-20 20:45:57 -04:00
|
|
|
return ret;
|
2011-02-01 15:38:37 -05:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-02-27 16:36:53 -05:00
|
|
|
static int download(const char *url, const char *localpath,
|
2011-02-10 11:19:41 -05:00
|
|
|
int force)
|
|
|
|
{
|
2009-04-04 04:17:30 -04:00
|
|
|
if(handle->fetchcb == NULL) {
|
2011-02-08 21:05:53 -05:00
|
|
|
#ifdef HAVE_LIBCURL
|
2011-03-20 20:45:57 -04:00
|
|
|
return curl_download_internal(url, localpath, force);
|
2008-04-06 22:00:11 -04:00
|
|
|
#else
|
|
|
|
RET_ERR(PM_ERR_EXTERNAL_DOWNLOAD, -1);
|
|
|
|
#endif
|
2008-02-27 16:36:53 -05:00
|
|
|
} else {
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
int ret = handle->fetchcb(url, localpath, force);
|
2009-04-04 04:17:30 -04:00
|
|
|
if(ret == -1) {
|
|
|
|
RET_ERR(PM_ERR_EXTERNAL_DOWNLOAD, -1);
|
|
|
|
}
|
2011-03-20 20:45:57 -04:00
|
|
|
return ret;
|
2008-02-27 16:36:53 -05:00
|
|
|
}
|
|
|
|
}
|
2008-01-22 23:00:12 -05:00
|
|
|
|
2008-01-22 22:36:11 -05:00
|
|
|
/*
|
2008-02-27 16:36:53 -05:00
|
|
|
* Download a single file
|
2008-08-03 19:16:08 -04:00
|
|
|
* - servers must be a list of urls WITHOUT trailing slashes.
|
2008-01-22 22:36:11 -05:00
|
|
|
*
|
|
|
|
* RETURN: 0 for successful download
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
* 1 if the files are identical
|
2008-01-22 22:36:11 -05:00
|
|
|
* -1 on error
|
|
|
|
*/
|
2008-02-27 16:36:53 -05:00
|
|
|
int _alpm_download_single_file(const char *filename,
|
|
|
|
alpm_list_t *servers, const char *localpath,
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
int force)
|
2008-01-22 22:36:11 -05:00
|
|
|
{
|
|
|
|
alpm_list_t *i;
|
2008-02-27 11:45:01 -05:00
|
|
|
int ret = -1;
|
2008-01-22 22:36:11 -05:00
|
|
|
|
2009-06-06 12:03:29 -04:00
|
|
|
ASSERT(servers != NULL, RET_ERR(PM_ERR_SERVER_NONE, -1));
|
|
|
|
|
2008-02-27 11:45:01 -05:00
|
|
|
for(i = servers; i; i = i->next) {
|
2008-01-22 23:49:45 -05:00
|
|
|
const char *server = i->data;
|
2008-02-27 16:36:53 -05:00
|
|
|
char *fileurl = NULL;
|
2009-10-11 15:02:20 -04:00
|
|
|
size_t len;
|
2008-02-27 16:36:53 -05:00
|
|
|
|
|
|
|
/* print server + filename into a buffer */
|
|
|
|
len = strlen(server) + strlen(filename) + 2;
|
|
|
|
CALLOC(fileurl, len, sizeof(char), RET_ERR(PM_ERR_MEMORY, -1));
|
|
|
|
snprintf(fileurl, len, "%s/%s", server, filename);
|
|
|
|
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
ret = download(fileurl, localpath, force);
|
2008-02-27 16:36:53 -05:00
|
|
|
FREE(fileurl);
|
|
|
|
if(ret != -1) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2008-01-22 22:36:11 -05:00
|
|
|
|
2011-03-20 20:45:57 -04:00
|
|
|
return ret;
|
2008-02-27 16:36:53 -05:00
|
|
|
}
|
2008-01-22 22:36:11 -05:00
|
|
|
|
2008-02-27 16:36:53 -05:00
|
|
|
int _alpm_download_files(alpm_list_t *files,
|
|
|
|
alpm_list_t *servers, const char *localpath)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
alpm_list_t *lp;
|
2008-01-22 22:36:11 -05:00
|
|
|
|
2008-02-27 16:36:53 -05:00
|
|
|
for(lp = files; lp; lp = lp->next) {
|
|
|
|
char *filename = lp->data;
|
|
|
|
if(_alpm_download_single_file(filename, servers,
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
localpath, 0) == -1) {
|
2008-02-27 16:36:53 -05:00
|
|
|
ret++;
|
2008-01-22 22:36:11 -05:00
|
|
|
}
|
|
|
|
}
|
2008-02-27 11:45:01 -05:00
|
|
|
|
2011-03-20 20:45:57 -04:00
|
|
|
return ret;
|
2008-01-22 22:36:11 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
/** Fetch a remote pkg.
|
|
|
|
* @param url URL of the package to download
|
|
|
|
* @return the downloaded filepath on success, NULL on error
|
|
|
|
* @addtogroup alpm_misc
|
|
|
|
*/
|
|
|
|
char SYMEXPORT *alpm_fetch_pkgurl(const char *url)
|
|
|
|
{
|
|
|
|
char *filename, *filepath;
|
|
|
|
const char *cachedir;
|
2008-02-27 16:36:53 -05:00
|
|
|
int ret;
|
2008-01-22 22:36:11 -05:00
|
|
|
|
|
|
|
ALPM_LOG_FUNC;
|
|
|
|
|
2008-02-27 16:36:53 -05:00
|
|
|
filename = get_filename(url);
|
2008-01-22 22:36:11 -05:00
|
|
|
|
|
|
|
/* find a valid cache dir to download to */
|
|
|
|
cachedir = _alpm_filecache_setup();
|
|
|
|
|
|
|
|
/* download the file */
|
download: major refactor to address lingering issues
Sorry for this being such a huge patch, but I believe it is necessary for
quite a few reasons which I will attempt to explain herein. I've been
mulling this over for a while, but wasn't super happy with making the
download interface more complex. Instead, if we carefully order things in
the internal download code, we can actually make the interface simpler.
1. FS#15657 - This involves `name.db.tar.gz.part` files being left around the
filesystem, and then causing all sorts of issues when someone attempts to
rerun the operation they canceled. We need to ensure that if we resume a
download, we are resuming it on exactly the same file; if we cannot be
almost postive of that then we need to start over.
2. http://www.mail-archive.com/pacman-dev@archlinux.org/msg03536.html - Here
we have a lighttpd bug to ruin the day. If we send both a Range: header and
If-Modified-Since: header across the wire in a GET request, lighttpd doesn't
do what we want in several cases. If the file hadn't been modified, it
returns a '304 Not Modified' instead of a '206 Partial Content'. We need to
do a stat (e.g. HEAD in HTTP terms) operation here, and the proceed
accordingly based off the values we get back from it.
3. The mtime stuff was rather ugly, and relied on the called function to
write back to a passed in reference, which isn't the greatest. Instead, use
the power of the filesystem to contain this info. Every file downloaded
internally is now carefully timestamped with the remote file time. This
should allow the resume logic to work. In order to guarantee this, we need
to implement a signal handler that catches interrupts, notifies the running
code, and causes it to set the mtimes on the file. It then rethrows the
signal so the pacman signal handler (or any frontend) works as expected.
4. We did a lot of funky stuff in trying to track the DB last modified time.
It is a lot easier to just keep the downloaded DB file around and track the
time on that rather than in a funky dot file. It also kills a lot of code.
5. For GPG verification of the databases down the road, we are going to need
the DB file around for at least a short bit of time anyway, so this gets us
closer to that.
Signed-off-by: Dan McGee <dan@archlinux.org>
[Xav: fixed printf with off_t]
Signed-off-by: Xavier Chantry <shiningxc@gmail.com>
2009-11-12 00:39:26 -05:00
|
|
|
ret = download(url, cachedir, 0);
|
2008-02-27 16:36:53 -05:00
|
|
|
if(ret == -1) {
|
2008-01-22 22:36:11 -05:00
|
|
|
_alpm_log(PM_LOG_WARNING, _("failed to download %s\n"), url);
|
2011-03-20 20:45:57 -04:00
|
|
|
return NULL;
|
2008-01-22 22:36:11 -05:00
|
|
|
}
|
2008-01-22 23:49:45 -05:00
|
|
|
_alpm_log(PM_LOG_DEBUG, "successfully downloaded %s\n", url);
|
2008-01-22 22:36:11 -05:00
|
|
|
|
|
|
|
/* we should be able to find the file the second time around */
|
|
|
|
filepath = _alpm_filecache_find(filename);
|
2011-03-20 20:45:57 -04:00
|
|
|
return filepath;
|
2008-01-22 22:36:11 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
/* vim: set ts=2 sw=2 noet: */
|