Exclude files with hardlinks when cat'ing all the files, and do a second
run to look at each file with hardlinks, keep track of the ones we've
already operated on, and only cat each inode once. Then use "wc -c" to get
the size of all (deduplicated) files the same way we were already doing.
Original-patch-by: Ronan Pigott <rpigott@berkeley.edu>
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
zipman:
read -r protects against those evil manpages whose filenames contain
backslash escapes, (muahahaha?)
IFS= read protects against filenames with:
- leading whitespace (but no one is actually stupid enough to configure
their MAN_DIRS=() in makepkg.conf with such silly directories, *right*?)
- trailing whitespace (but likewise, no one should be stupid enough to
write an uncompressed manpage for section '1 ' or something)
Also fix several other cases where we read filenames without protecting
against surrounding whitespace, or without using null-delimited
filenames when we could trivially do so.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Saving fflages breaks reproducible builds due to encoding information
specific to the filesystem that was used to build the package. This
information is not needed for packaging purposes anyway.
Including fflags also means that attempting to extract a package file as
root (or fakeroot) might result in angry warnings being printed to the
console by bsdtar, followed by a non-zero exit code, unless the user
remembers to use --no-fflags during extraction. This is unpleasant UI, even
if pacman itself won't care about these.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
remove_deps is called once, at the end of clean_up() before makepkg
exit. If remove_deps returns >0 (e.g. when pressing "n" in the resulting
prompt), the error is caught by the ERR signal handler. This in turns
sends SIGUSR1 to the process group, with resulting exit code 138.
In case remove_deps fails, this patch exits makepkg with E_REMOVE_DEPS
if there was no previous error (that is, EXIT_CODE equals E_OK).
Otherwise, makepkg exits with EXIT_CODE.
Signed-off-by: Allan McRae <allan@archlinux.org>
When running `makepkg -i` it may be necessary to first remove make- and
checkdepends before installing the built package - for example if they
conflict each other. This is the case for wireguard-arch which
makedepends and conflicts wireguard-dkms.
Signed-off-by: Erich Eckner <git@eckner.net>
Signed-off-by: Allan McRae <allan@archlinux.org>
Without the -f option to wait, we might move on and try to delete the
logpipe before the process is completed.
Signed-off-by: Allan McRae <allan@archlinux.org>
The logpipe fifo can remain when exiting on a non-error condition such
as recieving signals INT and USR1. This can be seen by doing either a
manual CTRL-C to interrupt the build or by sending a signal such as:
$ makepkg & sleep 5 ; kill -USR1 $!
Remove the fifo in all cases on script exit if it still exists.
Signed-off-by: Austin Lund <austin.lund@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
When scripting/automating around makepkg, it is sometimes desirable to
know how makepkg will be configured to operate. One example is the
archlinux devtools, which must forward select makepkg.conf variables
into a build chroot (for example PACKAGER) or use those variables itself
(for example {SRC,PKG,LOG}DEST).
The configuration file can be in up to 3 places, and should be capable
of being overridden via environment variables. It is sufficiently
complex to represent distinct functionality, and sufficiently useful to
merit easy accessibility in other scripts, therefore, let us move it
into a publicly exposed utility library.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Fixes "arch" and "checkdepends" never having been unset, fixes b2sums
(but not ${!b2sums_@}) being recently left out.
The "build" function used to be unset as well, explicitly unset it as a
function and do the same for other official functions as well.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
In commit b5191ea140 we moved to using
shell globbing to print package files for a couple of reasons including
reproducible packaging of .METADATA files.
Unfortunately, this only works reliably when the glob pattern does not
resolve to a symlinked directory due to a change in the bash 5.0
release. Note that the previous, desired behavior was rather to merely
refuse to recurse into symlinked directories, but due to an unrelated
issue, the symlink handling for globstar was reworked in a way that had
this side effect.
See https://lists.gnu.org/archive/html/bug-bash/2019-04/msg00015.html
for discussion; this may be fixed at some point, but bash 5.0 is broken
either way.
The appropriate way of handling this seems to be to use **/* to match
instead; this produces the same results on both bash 4 and bash 5, as
the ** matches any leading directory component (or none), and the *
matches any file, directory, or symlink to either one.
Fixes FS#62278
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Makepkg used to use du --apparent-size to compute the size of the
package. Unfortunately, this would result in different sizes depending
on the filesystem used (e.g., btrfs vs ext4), which would affect
reproducible builds. Use a wc-based approach to compute sizes
Signed-off-by: Santiago Torres <santiago@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
coreutils 8.26 in December 2016 added this new hashing method which is
compatible with the existing md5sum and sha*sum tool usage, while using
the blake2 hash algorithm.
makepkg uses coreutils to provide source file integrity checks via
${integ}sum binaries and it makes sense to offer this as an additional
option.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
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 61ba5c961e.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
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@compiletoi.net>
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Refactor many of the different arrays of pkgbuild variables
into scripts/libmakepkg/util/schema.sh.in.
Signed-off-by: morganamilo <morganamilo@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
This patch was inspired by FS#32723 which asks makepkg to install makedepends
before depends. The use case is to build a package depending on a virtual
package that is only provided by other packages (e.g. java-runtime in Arch
Linux), but wanting to build against a specific version. Installing makedepends
first (but not at the same time as depends) would allow specifying the version
to build against, instead of pacman resolving to the default version when
installing depends.
It turns out, we can already achieve installing makedepends first by specifying
dependencies only in the package function (and making sure makedepends includes
everything needed). The only issue is that if we use makepkg to install the
built package with the --install flag and along with the --rmdeps flag, we will
try to remove any installed dependencies that are specified in the depends
array in the package function. To counter this, we need to use the --unneeded
flag for the pacman call.
Signed-off-by: Allan McRae <allan@archlinux.org>
This is not really an error with a "user function".
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
paths can contain printf-unsafe chars, and printf -v is not somehow immune to this
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
There are state variables for everything else, and we use them to do
conditional checks on things, but it's currently a bit difficult to test
whether a package is being built, as it's the default action if *no*
options are specified.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Merge the similar code handling unsplit PKGBUILDs and individual
packages in a split PKGBUILD and make it a new function.
Signed-off-by: Allan McRae <allan@archlinux.org>
This opens the door for third parties to provide libmakepkg
extentions for the purpose of altering the build environment.
Signed-off-by: Que Quotion <quequotion@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
This opens the door for third parties who provide extensions to
libmakepkg to supply scripts that confirm the presence of their
dependant executables.
Signed-off-by: Que Quotion <quequotion@gmail.com>
Signed-off-by: Allan McRae <allan@archlinux.org>
If a user has a makepkg.conf policy to enable debug builds, but a
PKGBUILD has disabled buildflags, we would unset the *FLAGS but then
later append the debug *FLAGS anyway, which would result in some *FLAGS
being used, against the wishes of the PKGBUILD author.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
The respective write_* functions are low-level and shouldn't be
outputting statuses; move these to the logic flow where they are used.
This ensures the functions can be used in the future wherever, and also
solves an issue where, as fallout from the message.sh retrofitting in
commit 882e707e40, the statuses got
redirected to the actual files.
The resulting package was technically correct, except that it contained
useless lines which pacman ignored, and repo-add also ignored but at the
same time generated an error message:
/usr/bin/repo-add: line 335: declare: `=-> Generating .PKGINFO file...': not a valid identifier
Thirdparty package tools with stricter parsers may abort with errors,
and "repose" is known to do so.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This behavior is confusing, since it means absolutely everything goes to
stderr and makepkg itself is a quiet program that produces no expected
output???
The only situation where messages should go to stderr rather than
stdout, is with --geninteg which is meant to return the checksums on
stdout (but we don't want to totally get rid of status messages when
redirecting the results elsewhere, or, worse, redirect status messages
to a PKGBUILD). For this specific case, redirect message output to
stderr in the --geninteg callers directly.
Implements FS#17173
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
- it comes with free collation when moving the LC_ALL declaration up a bit;
this fixes a bug where the .FILES were not being properly sorted and
their order depended on directory creation order, which broke
reproducible builds in the wild.
- it handles sorting null-delimited output everywhere, without sort -z;
this lets us get rid of sed hacks
- it is faster than invoking multiple find subprocesses
- dotfiles can be automatically printed *and the C locale sorts them first*
with a single ** glob
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
We don't need to translate the "Copyright YEAR AUTHOR" part, no part of
it should probably be translated and it definitely shouldn't turn every
single license terms notice into a separate translation just because the
author/year is different.
Fixes FS#58452
Also consistently add a blank line after the copyright and before the
license terms.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This reverts commit 9cdfd18739.
We've never documented whirlpoolsums support in the manpage and no one
really seems to have realized we support it, let alone use it -- except
for a few parabola packages, being the contributor's motivation for
adding support.
The problem is that for two years the code has been broken. In commit
577701250d we moved to coreutils to
provide checksum commands, rather than openssl, but there is no
whirlpoolsums binary.
Properly fixing this would require re-adding a dependency on openssl,
independent of the libalpm crypto backend -- which defeats the purpose
of moving to coreutils in the general case. nettle-hash does not provide
a whirlpool algorithm any more than it does base64 (the original reason
for moving to coreutils).
Therefore, we should just drop support for this again.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
It's most likely a case where output is being captured, so we shouldn't
be interleaving status messages with function output regardless. Setting
the pkgver() status message (the one time we use it in a subshell)
separately also makes it safe to change whether message.sh functions write
to stdout or stderr.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
`run_function_safe pkgver` is evaluated in a subshell and therefore does
not abort when it should. Explicitly check the return outside of the
subshell and abort if necessary.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Both run_function and run_function_safe will save and restore `shopt -p`
but the former is only called from the latter. It makes sense to save
this as part of a "safe" runner, so let's just do it in one place, there
where we save and restore everything else too.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
When re-running makepkg for fakeroot, if `bash -x makepkg` was used this
is lost. Fix by encoding the current set of options explicitly in the
invocation, both for makepkg and for the wrapper used to test scripts
inside the source tree.
Also change to use ${BASH_SOURCE[0]} instead of $0 as the latter can be
anything the parent process wants, while the former is explicitly set by
bash itself to the filepath of the script.
See http://mywiki.wooledge.org/BashFAQ/028
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This causes package_$pkgname() to be preferred over package() in the
non-split case, but the behavior if both functions exist was
undocumented anyway.
Signed-off-by: Allan McRae <allan@archlinux.org>
The biggest issue is directly supplying the data within the format
string which can result in misinterpreting formatter sequences if a
printed variable contains an "%" in it. This character is currently
permitted in the pkgver field, though not in the pkgname. Also
pacman/libalpm itself has much looser limitations and this can appear
anywhere at all if a package was created by some other program.
For the package "iambroke-1%s-1-any.pkg.tar.xz", installed in the build
environment, the result is:
-> Generating .BUILDINFO file...
awk: cmd. line:3: (FILENAME=- FNR=1085) fatal: not enough arguments to satisfy format string
`-1%s-1'
^ ran out for this one
Followed by a .BUILDINFO which contains an LC_ALL=C sorted list of
$pkgname-${epoch:+$epoch:}$pkgver-$pkgrel-$arch ending in:
installed = iambroke
Which is cut short, then fails to list the succeeding packages. The
package itself successfully builds.
It's also unnecessary to save the output of pacman -Qq in order to get the
information for pacman -Qi, since the latter will, just like the former,
return information for all installed packages if not given a package
name(s).
While I am at it, pipe this directly to awk rather than keeping a copy
in an unnecessary local variable. This is slightly more efficient in
addition to preventing the <<< herestring from re-interpreting the
content of "$pkginfos" in ways that don't really matter for our usage.
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
This reverts commit 9e52a36794.
The change to use declare for the split package metadata backup/restore
resulted in variables being declared at a local scope. When these variables
were unset (mostly noticed with debug packaging) this left the variable at
global scope defined. Revert back to the known good state.
Signed-off-by: Allan McRae <allan@archlinux.org>
When comparing the $BUILDDIR to the $startdir, we used string equality
instead of testing whether they are the same location, and ended up
appending $pkgbase even though there's no reason to use it here.
In some cases, this could result in makepkg erroring when trying to
create $srcdir/$pkgdir, if a file with the same name as the $pkgbase
exists. This is expected behavior if a file "src" or "pkg" exists, but
decidedly less so for $pkgbase.
This could be fixed either by setting $startdir to an absolute path, or
by ensuring the test checks this directly; I've chosen to do both, since
the test should really be correctly checking the thing it actually cares
about, but since we ensure absolute paths are used everywhere else, this
might bite us elsewhere someday. It's also more consistent.
Fixes FS#58865
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
It's especially dangerous in trap handlers since the return value of the
function becomes the return value of the last command before the trap,
not the last command in the current function. This applies to any
function executed in a trap handler, nested functions included.
In one case, install_packages failed (via return 14), which was inside a
conditional that then ran exit 14, which triggered the EXIT handler,
which called clean_up, which called remove_deps, which had !RMDEPS and
thus returned. The return value of remove_deps became the return value
of install_packages, triggering the ERR handler, which (due to another
problem) was still the user function handler, which then printed a
misleading error message and overrode the exit code with 4.
Signed-off-by: Allan McRae <allan@archlinux.org>
In commit 8ff03868a3 PACMAN_OPTS was
turned into an array. Unfortunately, that array was generated by
treating the "--color never" option as one string, instead of an
array of two strings...
Fixes FS#58820
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
$restoretrap is empty if the trap was not set. This caused the trap
handler to remain and override later exit codes.
Signed-off-by: Allan McRae <allan@archlinux.org>
In commit d8717a6a96 the write permission
checks were refactored. Initially we intended to drop this chmod in the
process, but due to some confusion about whether it was needed, I ended
up submitting patches both to preserve and to remove it... but it's not
needed after all. We do it on the individual $srcdir/$pkgdir, later on.
Then, we used the wrong version, which causes unnecessary restrictions.
See FS#58790
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
Reported-by: Rafael Ascensão <rafa.almas@gmail.com>
Signed-off-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Allan McRae <allan@archlinux.org>
