Attachment 9537 Details for Bug 40582 – create_chroot(): pin %%_rpmscript_werror (as expected in Girar install checks)

[patch] create_chroot(): pin %%_rpmscript_werror (as expected in Girar install checks)

0001-create_chroot-pin-_rpmscript_werror-as-expected-in-G.patch (text/plain), 2.45 KB, created by Ivan Zakharyaschev on 2021-07-26 21:40:24 MSK

(hide)

Description:

Filename:

MIME Type:

Creator: Ivan Zakharyaschev

Created: 2021-07-26 21:40:24 MSK

Size: 2.45 KB

patch

obsolete

>From 8c3acab99cdd866f9fc978bc6cf685d198d4d6cd Mon Sep 17 00:00:00 2001
>From: Ivan Zakharyaschev <imz@altlinux.org>
>Date: Mon, 26 Jul 2021 17:00:27 +0300
>Subject: [PATCH] create_chroot(): pin %%_rpmscript_werror (as expected in
> Girar install checks)
>
>(I don't like hardcoding this; a more flexible solution allowing to
>choose the behavior for each invocation of hsh --ini or hsh-install
>would be better: it could be an option to pass the corresponding
>option along with rpmi.)
>
>(For this change to take effect, one should rm -rf ~/hasher/cache/chroot/ )
>
>For non-zero exit codes of the %%post scriplets to be treated as fatal
>errors (in install checks in Girar), %%_rpmscript_werror flag was recently
>added to rpm, and an attempt was made to turn this behavior on in
>Girar in commit 6c5d2738e0ad04d00c31c980a6e8f71ff5bb24ab:
>
>Author: Dmitry V. Levin <ldv@altlinux.org>
>Date:   Thu Dec 19 00:44:59 2019 +0000
>
>    gb-remote-plant: add %_rpmscript_werror to .rpmmacros
>
>    Enable traditional rpm behaviour of treating all scriptlet failures
>    as fatal.
>
>However, adding it to ~/.rpmmacros didn't have the desired effect,
>because, in hasher, firstly, rpmmacros from the host system are
>primarily used to fill in builder's rpmmacros (and affect only
>building packages), and secondly, even there they are not copied
>verbatim, but rather some needed values are evaluated and
>inserted. (See "# Create RPM build directory tree." in hsh-initroot.)
>Also, aptbox sees ~/.rpmmacros from the host system (or re-creates its
>own $aptbox_home/.rpmmacros), but they don't seem to affect the
>behavior of rpm w.r.t. the exit status of the scriptlets of the
>packages being installed.
>
>This is demonstrated by the trivial test packages from
>http://git.altlinux.org/people/imz/packages/rpmscript_warn_only-checkinstall.git
>
>Reported-by: Alexey Appolonov <alexey@altlinux.org>
>---
> hasher/hsh-sh-cache-chroot-functions | 2 ++
> 1 file changed, 2 insertions(+)
>
>diff --git a/hasher/hsh-sh-cache-chroot-functions b/hasher/hsh-sh-cache-chroot-functions
>index 34a5b95..1efffa5 100644
>--- a/hasher/hsh-sh-cache-chroot-functions
>+++ b/hasher/hsh-sh-cache-chroot-functions
>@@ -277,6 +277,8 @@ mkdir -p /sys
> rm -f /etc/rpm/macros.db1
> # WORKAROUND: glibc-locales is too large
> echo '%_install_langs ${install_langs:-$def_install_langs}' >>/etc/rpm/macros
>+# Force the treatment of scriptlet errors as expected in install checks in Girar
>+echo '%_rpmscript_werror 1' >>/etc/rpm/macros
> # /WORKAROUND
> rpmdb --initdb
> __EOF__
>-- 
>2.25.4
>

Actions: View | Diff

Attachments on bug 40582: 9537