February 2019 - Pacman-dev - lists.archlinux.org

[pacman-dev] [PATCH] makepkg: simplify run_pacman logic
by Eli Schwartz 21 Feb '19

21 Feb '19

We don't need to check the options twice, since it is the same check both times. Instead, merge the conditionals. As far as I can tell, the only reason the checks for: - PACMAN_OPTS and - whether to use sudo were ever separated is due to the historic existence of --asroot, since the second check included a check for (( ! ASROOT )) until it was cleaned up in commit 61ba5c961e4a3536c4bbf41edb348987a9993fdb. Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> --- scripts/makepkg.sh.in | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in index 25700b09..c51fec45 100644 --- a/scripts/makepkg.sh.in +++ b/scripts/makepkg.sh.in @@ -217,12 +217,10 @@ missing_source_file() { run_pacman() { local cmd - if [[ $1 != -@(T|Q)*([[:alpha:]]) ]]; then - cmd=("$PACMAN_PATH" "${PACMAN_OPTS[@]}" "$@") - else + if [[ $1 = -@(T|Q)*([[:alpha:]]) ]]; then cmd=("$PACMAN_PATH" "$@") - fi - if [[ $1 != -@(T|Q)*([[:alpha:]]) ]]; then + else + cmd=("$PACMAN_PATH" "${PACMAN_OPTS[@]}" "$@") if type -p sudo >/dev/null; then cmd=(sudo "${cmd[@]}") else -- 2.20.1

2 1

[pacman-dev] [PATCH] makepkg: implement locking for pacman commands
by Eli Schwartz 21 Feb '19

21 Feb '19

When pacman is run as root to do -S, -U, or -R, it would immediately abort if pacman is not ready for use. Instead, poll the lockfile and wait until it becomes available. Implements FS#28840 Original-patch-by: Georges Dubus <georges.dubus(a)compiletoi.net> Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> --- scripts/makepkg.sh.in | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in index cf30c302..25700b09 100644 --- a/scripts/makepkg.sh.in +++ b/scripts/makepkg.sh.in @@ -228,6 +228,14 @@ run_pacman() { else cmd=(su root -c "$(printf '%q ' "${cmd[@]}")") fi + + local lockfile="$(pacman-conf DBPath)/db.lck" + if [[ -f $lockfile ]]; then + msg "$(gettext "Pacman is currently in use, please wait...")" + while [[ -f $lockfile ]]; do + sleep 3 + done + fi fi "${cmd[@]}" } -- 2.20.1

2 2

[pacman-dev] [GIT] The official pacman repository branch, master, updated. v5.1.1-134-g477a66cd
by Allan McRae 12 Feb '19

12 Feb '19

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "The official pacman repository". The branch, master has been updated via 477a66cd0e7a1b8f5fb7d3b79d977da2771ee09a (commit) via 86004227d14bef059f71290b3c0454ad88ea0f86 (commit) from db102c67efe9cfd4367a9110c8f97ceff5afcbe3 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 477a66cd0e7a1b8f5fb7d3b79d977da2771ee09a Author: Eli Schwartz <eschwartz(a)archlinux.org> Date: Mon Feb 11 11:19:26 2019 -0500 build: link vercmp with a static copy of libalpm This has historically been the case in autotools since we want vercmp to not break mid-transaction in an install script. For convenience, we create libalpm.a and use this to optionally generate libalpm.so (when not configured with -Dbuildstatic=true) as well as to link any binary which explicitly wishes to be built statically "with libalpm", but does not care where a function is defined. meson then treats this correctly: it builds the object file only once for both libraries, and the compiler strips out unused functionality from the final static binary. Currently the only binary which requires this is vercmp. Fixes FS#61719 Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> Signed-off-by: Allan McRae <allan(a)archlinux.org> commit 86004227d14bef059f71290b3c0454ad88ea0f86 Author: Allan McRae <allan(a)archlinux.org> Date: Tue Feb 12 11:17:30 2019 +1000 Add implicit fall through warning Requires modification to our comment about fall through to match compilers expectations. Works for GCC and Clang. Signed-off-by: Allan McRae <allan(a)archlinux.org> ----------------------------------------------------------------------- Summary of changes: configure.ac | 1 + meson.build | 16 +++++++++++++--- src/pacman/pacman.c | 2 +- 3 files changed, 15 insertions(+), 4 deletions(-) hooks/post-receive -- The official pacman repository

1 0

[pacman-dev] [PATCH] build: link vercmp with a static copy of libalpm
by Eli Schwartz 12 Feb '19

12 Feb '19

This has historically been the case in autotools since we want vercmp to not break mid-transaction in an install script. For convenience, we create libalpm.a and use this to optionally generate libalpm.so (when not configured with -Dbuildstatic=true) as well as to link any binary which explicitly wishes to be built statically "with libalpm", but does not care where a function is defined. meson then treats this correctly: it builds the object file only once for both libraries, and the compiler strips out unused functionality from the final static binary. Currently the only binary which requires this is vercmp. Fixes FS#61719 Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> --- This is based on my initial suggestion via IRC for how to solve the vercmp build issue using two libalpm libraries. The important distinction is that in order to not be an error when buildstatic is used, we cannot try to create multiple rules that generate libalpm.a -- and also we don't want to install /usr/lib/libalpm_a.a meson.build | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/meson.build b/meson.build index 0a710653..4a74786b 100644 --- a/meson.build +++ b/meson.build @@ -354,15 +354,24 @@ libcommon = static_library( include_directories : includes, install : false) -libalpm = library( +libalpm_a = static_library( 'alpm', libalpm_sources, - version : libalpm_version, include_directories : includes, dependencies : [crypto_provider, libarchive, libcurl] + gpgme_libs, link_with : [libcommon], install : true) +if not get_option('buildstatic') + libalpm = shared_library( + 'alpm', + version : libalpm_version, + link_whole: [libalpm_a], + install : true) +else + libalpm = libalpm_a +endif + install_headers( 'lib/libalpm/alpm.h', 'lib/libalpm/alpm_list.h') @@ -413,7 +422,7 @@ executable( 'vercmp', vercmp_sources, include_directories : includes, - link_with : [libalpm], + link_with : [libalpm_a], install : true, ) -- 2.20.1

2 1

[pacman-dev] [PATCH] Add implicit fall through warning
by Allan McRae 12 Feb '19

12 Feb '19

Requires modification to our comment about fall through to match compilers expectations. Works for GCC and Clang. Signed-off-by: Allan McRae <allan(a)archlinux.org> --- configure.ac | 1 + meson.build | 1 + src/pacman/pacman.c | 2 +- 3 files changed, 3 insertions(+), 1 deletion(-) diff --git a/configure.ac b/configure.ac index 415ed3cb..2f345b5d 100644 --- a/configure.ac +++ b/configure.ac @@ -462,6 +462,7 @@ if test "x$warningflags" = "xyes" ; then CFLAGS_ADD([-Wformat-nonliteral], [WARNING_CFLAGS]) CFLAGS_ADD([-Wformat-security], [WARNING_CFLAGS]) CFLAGS_ADD([-Wignored-qualifiers], [WARNING_CFLAGS]) + CFLAGS_ADD([-Wimplicit-fallthrough], [WARNING_CFLAGS]) CFLAGS_ADD([-Winit-self], [WARNING_CFLAGS]) CFLAGS_ADD([-Wlogical-op], [WARNING_CFLAGS]) CFLAGS_ADD([-Wmissing-declarations], [WARNING_CFLAGS]) diff --git a/meson.build b/meson.build index 0a710653..02a3a3d3 100644 --- a/meson.build +++ b/meson.build @@ -220,6 +220,7 @@ if get_option('buildtype').startswith('debug') '-Wformat-nonliteral', '-Wformat-security', '-Wignored-qualifiers', + '-Wimplicit-fallthrough', '-Winit-self', '-Wlogical-op', '-Wmissing-declarations', diff --git a/src/pacman/pacman.c b/src/pacman/pacman.c index a2a420b6..3bb622e6 100644 --- a/src/pacman/pacman.c +++ b/src/pacman/pacman.c @@ -196,7 +196,7 @@ static void usage(int op, const char * const myname) addlist(_(" --ignore <pkg> ignore a package upgrade (can be used more than once)\n")); addlist(_(" --ignoregroup <grp>\n" " ignore a group upgrade (can be used more than once)\n")); - /* pass through */ + /* fall through */ case PM_OP_REMOVE: addlist(_(" -d, --nodeps skip dependency version checks (-dd to skip all checks)\n")); addlist(_(" --assume-installed <package=version>\n" -- 2.20.1

1 0

[pacman-dev] [PATCH] build: link vercmp staticly with libalpm
by Dave Reisner 11 Feb '19

11 Feb '19

This makes the meson-built vercmp equivalent to the autotools build. Employ an intermediate archive of libalpm which our generic library rule can slurp in. Other alpm-only binaries (specifically we care about vercmp) can then link this in for alpm functionality without caring exactly where the function is defined. meson passes the right flags to ensure that unused data is stripped out of the executable. ref: FS#61719 --- meson.build | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/meson.build b/meson.build index 0a710653..5c1efd73 100644 --- a/meson.build +++ b/meson.build @@ -354,12 +354,21 @@ libcommon = static_library( include_directories : includes, install : false) +libalpm_a = static_library( + 'alpm_a', + libalpm_sources, + version : libalpm_version, + include_directories : includes, + dependencies : [crypto_provider, libarchive, libcurl] + gpgme_libs, + link_with : [libcommon], + install : true) + libalpm = library( 'alpm', - libalpm_sources, version : libalpm_version, include_directories : includes, dependencies : [crypto_provider, libarchive, libcurl] + gpgme_libs, + link_whole : [libalpm_a], link_with : [libcommon], install : true) @@ -413,7 +422,7 @@ executable( 'vercmp', vercmp_sources, include_directories : includes, - link_with : [libalpm], + link_with : [libalpm_a], install : true, ) -- 2.20.1

3 2

[pacman-dev] [GIT] The official pacman repository branch, master, updated. v5.1.1-132-gdb102c67
by Allan McRae 06 Feb '19

06 Feb '19

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "The official pacman repository". The branch, master has been updated via db102c67efe9cfd4367a9110c8f97ceff5afcbe3 (commit) via 714a1b4e4eaf7ffde5f30eccb476f59979fac44e (commit) from 04e77591d6333002815e9276ca08c01230fcf212 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit db102c67efe9cfd4367a9110c8f97ceff5afcbe3 Author: Mark Ulrich <mark.ulrich.86(a)gmail.com> Date: Wed Feb 6 08:36:51 2019 -0600 libalpm: prevent 301 redirect loop from hanging the process If a mirror responds with a 301 redirect to itself, it will create an infinite redirect loop. This will cause pacman to hang, unresponsive to even a SIGINT. The result is pacman being unable to sync or download any package from a particular repo if its current mirror is stuck in a redirect loop. Setting libcurl's MAXREDIRS option effectively prevents a redirect loop from hanging the process. Signed-off-by: Mark Ulrich <mark.ulrich.86(a)gmail.com> Signed-off-by: Allan McRae <allan(a)archlinux.org> commit 714a1b4e4eaf7ffde5f30eccb476f59979fac44e Author: Allan McRae <allan(a)archlinux.org> Date: Mon Feb 4 10:32:50 2019 +1000 I blame everyone but me for this Signed-off-by: Allan McRae <allan(a)archlinux.org> ----------------------------------------------------------------------- Summary of changes: NEWS | 2 +- lib/libalpm/dload.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) hooks/post-receive -- The official pacman repository

1 0

[pacman-dev] [PATCH] libalpm: prevent 301 redirect loop from hanging the process
by Mark Ulrich 06 Feb '19

06 Feb '19

If a mirror responds with a 301 redirect to itself, it will create an infinite redirect loop. This will cause pacman to hang, unresponsive to even a SIGINT. The result is pacman being unable to sync or download any package from a particular repo if its current mirror is stuck in a redirect loop. Setting libcurl's MAXREDIRS option effectively prevents a redirect loop from hanging the process. Signed-off-by: Mark Ulrich <mark.ulrich.86(a)gmail.com> --- lib/libalpm/dload.c | 1 + 1 file changed, 1 insertion(+) diff --git a/lib/libalpm/dload.c b/lib/libalpm/dload.c index 36ae4ee1..d04a5e46 100644 --- a/lib/libalpm/dload.c +++ b/lib/libalpm/dload.c @@ -259,6 +259,7 @@ static void curl_set_handle_opts(struct dload_payload *payload, curl_easy_setopt(curl, CURLOPT_URL, payload->fileurl); curl_easy_setopt(curl, CURLOPT_ERRORBUFFER, error_buffer); curl_easy_setopt(curl, CURLOPT_CONNECTTIMEOUT, 10L); + curl_easy_setopt(curl, CURLOPT_MAXREDIRS, 1L); curl_easy_setopt(curl, CURLOPT_FILETIME, 1L); curl_easy_setopt(curl, CURLOPT_NOPROGRESS, 0L); curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L); -- 2.20.1

3 4

[pacman-dev] Pacman failing with existing symlinks on the install path
by Peter Kehl 04 Feb '19

04 Feb '19

Dear Pacman maintainers, On current 4.19.16-1-MANJARO x64, I had /opt as symlink to /share/opt. I went to install current Google Chrome 72.0.3626.81-1 from https://aur.archlinux.org/packages/google-chrome (either by GNOME's Add/Remove Software, or by sudo pacman -U google-chrome-72.0.3626.81-1- x86_64.pkg.tar.xz). It failed with: error: failed to commit transaction (conflicting files) google-chrome: /opt exists in filesystem Knut Ahlers, the maintainer of AUR Google Chrome package, responded that the package doesn't check for the symlinks. Hence it's probably pacman. Have you/would you consider supporting existing symlinks on the install paths, please? Thanks in advance, -Peter Kehl Vancouver, B.C., Canada

3 3

[pacman-dev] [PATCH] libalpm: Check cachedir access before re-creating
by David Phillips 04 Feb '19

04 Feb '19

If users have mounted a filesystem such as sshfs or NFS that doesn't allow root access by default for security reasons, _alpm_filecache_setup will blindly try and re-create the components of the cachedir path, eating EACCES (to be liek mkdir -p). Then, it will return this path as if it is valid when it may not be readable by root, causing alpm_check_downloadspace to trigger a FPE since it assumes the cachedir is R/W and uses fsinfo which may be zero. This patch adds a check before trying to re-create the cachedir in order to determine if creating it will result in EACCES, and outputs a warning if so. It also discards the current cachedir from being considered. This causes a more meaningful and graceful failure than a FPE. --- lib/libalpm/util.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/lib/libalpm/util.c b/lib/libalpm/util.c index d33eef2a..d4375593 100644 --- a/lib/libalpm/util.c +++ b/lib/libalpm/util.c @@ -849,12 +849,16 @@ const char *_alpm_filecache_setup(alpm_handle_t *handle) for(i = handle->cachedirs; i; i = i->next) { cachedir = i->data; if(stat(cachedir, &buf) != 0) { - /* cache directory does not exist.... try creating it */ - _alpm_log(handle, ALPM_LOG_WARNING, _("no %s cache exists, creating...\n"), - cachedir); - if(_alpm_makepath(cachedir) == 0) { - _alpm_log(handle, ALPM_LOG_DEBUG, "using cachedir: %s\n", cachedir); - return cachedir; + if (errno == EACCES) { + _alpm_log(handle, ALPM_LOG_WARNING, _("failed to stat %s: %s\n"), cachedir, strerror(errno)); + } else { + /* cache directory does not exist.... try creating it */ + _alpm_log(handle, ALPM_LOG_WARNING, _("no %s cache exists, creating...\n"), + cachedir); + if(_alpm_makepath(cachedir) == 0) { + _alpm_log(handle, ALPM_LOG_DEBUG, "using cachedir: %s\n", cachedir); + return cachedir; + } } } else if(!S_ISDIR(buf.st_mode)) { _alpm_log(handle, ALPM_LOG_DEBUG, -- 2.20.1

3 2