February 2021 - Pacman-dev - lists.archlinux.org

[pacman-dev] [PATCH] makepkg: Don't double-layer distcc on ccache
by Matti Niemenmaa 08 Feb '21

08 Feb '21

From: Matti Niemenmaa <matti.niemenmaa+git(a)iki.fi> buildenv is set once for build() and a second time for package(). When using both distcc and ccache, this lead to CCACHE_PREFIX="distcc distcc" in package(), which breaks PKGBUILDs that execute the compiler in package() because distcc complains: distcc[383041] (main) CRITICAL! distcc seems to have invoked itself recursively! Avoid causing this error by only adding "distcc" to CCACHE_PREFIX if it's not yet there. Signed-off-by: Matti Niemenmaa <matti.niemenmaa+git(a)iki.fi> --- scripts/libmakepkg/buildenv/compiler.sh.in | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) ncurses is an example of a package that runs into this problem. Setting buildenv twice was done on purpose in commit 02a0bf550a22e199f48537b7eee87361b112e8a0, but at a glance I'm not sure whether it provides any benefit. All the meaningful changes are to environment variables that are exported from the parent process anyway, so all the changes are either repeated or even duplicated (like in PATH, or this CCACHE_PREFIX issue). Maybe it would be better to simply remove the prepare_buildenv call from the INFAKEROOT case in makepkg.sh.in? diff --git a/scripts/libmakepkg/buildenv/compiler.sh.in b/scripts/libmakepkg/buildenv/compiler.sh.in index 69f58a29..c93c77b4 100644 --- a/scripts/libmakepkg/buildenv/compiler.sh.in +++ b/scripts/libmakepkg/buildenv/compiler.sh.in @@ -44,7 +44,9 @@ buildenv_ccache() { buildenv_distcc() { if check_buildoption "distcc" "y"; then if (( using_ccache )); then - export CCACHE_PREFIX="${CCACHE_PREFIX:+$CCACHE_PREFIX }distcc" + if ! [[ "$CCACHE_PREFIX" =~ (^| )distcc($| ) ]]; then + export CCACHE_PREFIX="${CCACHE_PREFIX:+$CCACHE_PREFIX }distcc" + fi export CCACHE_BASEDIR="$srcdir" elif [[ -d /usr/lib/distcc/bin ]]; then export PATH="/usr/lib/distcc/bin:$PATH" -- 2.24.1

3 6

[pacman-dev] [PATCH] doxygen: fix missing parameter name
by Eli Schwartz 08 Feb '21

08 Feb '21

This was raising a warning during the build. Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> --- lib/libalpm/alpm_list.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/lib/libalpm/alpm_list.h b/lib/libalpm/alpm_list.h index a0299a704..f76961d47 100644 --- a/lib/libalpm/alpm_list.h +++ b/lib/libalpm/alpm_list.h @@ -61,7 +61,7 @@ typedef struct __alpm_list_t { #define FREELIST(p) do { alpm_list_free_inner(p, free); alpm_list_free(p); p = NULL; } while(0) /** item deallocation callback. - * @param the item to free + * @param item the item to free */ typedef void (*alpm_list_fn_free)(void * item); -- 2.30.0

2 1

[pacman-dev] [PATCH 1/2] makepkg: don't let the strip program mess up file attributes
by Eli Schwartz 08 Feb '21

08 Feb '21

It updates the stripped file by creating a temp file, chown/chmodding it, and replacing the original file. But upstream binutils has CVE-worthy issues with this if running strip as root, and some recent versions of strip don't play nicely with fakeroot. Also, this has always destroyed xattrs. :/ Sidestep the issue by telling strip to write to a temporary file, and manually dump the contents of that back into the original binary. Since the original binary is intact, albeit with different contents, it retains its correct attributes in fakeroot. Signed-off-by: Eli Schwartz <eschwartz(a)archlinux.org> --- scripts/libmakepkg/tidy/strip.sh.in | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/scripts/libmakepkg/tidy/strip.sh.in b/scripts/libmakepkg/tidy/strip.sh.in index 4d50f4475..f7238f813 100644 --- a/scripts/libmakepkg/tidy/strip.sh.in +++ b/scripts/libmakepkg/tidy/strip.sh.in @@ -93,7 +93,10 @@ strip_file() { fi fi - strip $@ "$binary" + if strip "$@" "$binary" -o "$binary.stripped"; then + cat "$binary.stripped" > "$binary" + fi + rm -f "$binary.stripped" } -- 2.30.0

2 6

[pacman-dev] [PATCH] libmakepkg: maintain file ownership while stripping
by Allan McRae 07 Feb '21

07 Feb '21

fakeroot can be flakey, as demonstrated by strip not maintaining file permissions under fakeroot with binutils-2.36.1 built against glibc-2.33. Instead of relying on fakeroot to be functional, ensure permissions are maintained across file stripping. Also set debug symbols to always have root:root ownership. Signed-off-by: Allan McRae <allan(a)archlinux.org> --- build-aux/edit-script.sh.in | 1 + meson.build | 4 ++++ scripts/libmakepkg/tidy/strip.sh.in | 6 ++++++ 3 files changed, 11 insertions(+) diff --git a/build-aux/edit-script.sh.in b/build-aux/edit-script.sh.in index 661c22d5..fa5ebcf2 100644 --- a/build-aux/edit-script.sh.in +++ b/build-aux/edit-script.sh.in @@ -19,6 +19,7 @@ mode=$3 -e "s|@TEMPLATE_DIR[@]|@TEMPLATE_DIR@|g" \ -e "s|@DEBUGSUFFIX[@]|@DEBUGSUFFIX@|g" \ -e "s|@INODECMD[@]|@INODECMD@|g" \ + -e "s|@STATCMD[@]|@STATCMD@|g" \ -e "s|@FILECMD[@]|@FILECMD@|g" \ "$input" >"$output" diff --git a/meson.build b/meson.build index ea9bd2f7..edf988e9 100644 --- a/meson.build +++ b/meson.build @@ -224,6 +224,7 @@ add_project_arguments('-include', 'config.h', language : 'c') filecmd = 'file' inodecmd = 'stat -c \'%i %n\'' +statcmd = 'stat -c' strip_binaries = '--strip-all' strip_shared = '--strip-unneeded' strip_static = '--strip-debug' @@ -238,11 +239,13 @@ endif os = host_machine.system() if os.startswith('darwin') inodecmd = '/usr/bin/stat -f \'%i %N\'' + statcmd = '/usr/bin/stat -c' strip_binaries = '' strip_shared = '-s' strip_static = '-s' elif os.contains('bsd') or os == 'dragonfly' inodecmd = 'stat -f \'%i %N\'' + statcmd = 'stat -c' endif chost = run_command(cc, '-dumpmachine').stdout().strip() @@ -269,6 +272,7 @@ substs.set('BUILDSCRIPT', BUILDSCRIPT) substs.set('TEMPLATE_DIR', get_option('makepkg-template-dir')) substs.set('DEBUGSUFFIX', get_option('debug-suffix')) substs.set('INODECMD', inodecmd) +substs.set('STATCMD', statcmd) substs.set('FILECMD', filecmd) substs.set('LIBMAKEPKGDIR', LIBMAKEPKGDIR) substs.set('STRIP_BINARIES', strip_binaries) diff --git a/scripts/libmakepkg/tidy/strip.sh.in b/scripts/libmakepkg/tidy/strip.sh.in index 868b96f3..6676c012 100644 --- a/scripts/libmakepkg/tidy/strip.sh.in +++ b/scripts/libmakepkg/tidy/strip.sh.in @@ -66,10 +66,15 @@ strip_file() { fi done < <(source_files "$binary") + # save original file permissions + file_owner=$(@STATCMD@ '%u' "$binary") + file_group=$(@STATCMD@ '%g' "$binary") + # copy debug symbols to debug directory mkdir -p "$dbgdir/${binary%/*}" objcopy --only-keep-debug "$binary" "$dbgdir/$binary.debug" objcopy --add-gnu-debuglink="$dbgdir/${binary#/}.debug" "$binary" + chown root:root "$dbgdir/$binary.debug" # create any needed hardlinks while IFS= read -rd '' file ; do @@ -94,6 +99,7 @@ strip_file() { fi strip $@ "$binary" + chown $file_owner:$file_group "$binary" } -- 2.30.0

1 0

[pacman-dev] makepkg --packagelist while using --source or --allsource
by ＠RubenKelevra 03 Feb '21

03 Feb '21

Hey guys, I want to package some source files and discovered that --packagelist does not change it's output when using --source or --allsource. Since makepkg isn't generating the normal packages with these flags anyway, it's a bit confusing that the output isn't changing. It would be pretty helpful for my usecase to have makepkg printing the sourcepackage-filepaths instead of the regular filepaths. Example: [ruben@i3 brave-bin]$ makepkg --source --packagelist /home/ruben/tmp/makepkg_test/brave-bin/brave-bin-1:1.19.88-1-x86_64.pkg.tar.lz4 [ruben@i3 brave-bin]$ makepkg --packagelist /home/ruben/tmp/makepkg_test/brave-bin/brave-bin-1:1.19.88-1-x86_64.pkg.tar.lz4 Best regards, Ruben

2 2

[pacman-dev] makepkg default flags
by Christopher W. 03 Feb '21

03 Feb '21

A few months back there was a proposal to edit the default makepkg build flags for improved security: https://lists.archlinux.org/pipermail/arch-dev-public/2020-July/030031.html As far as I can tell, this was never committed. The show-stopper at the time was Clang 11 not being released yet. Now that it's been in the repos for a while, could we revisit this improvement and possibly get a mass rebuild with it applied?

2 1