On 13/01/21 07:48, Thomas Huth wrote:
[254/8314] Generating texture-blit-frag.h with a meson_exe.py custom
command
FAILED: ui/shader/texture-blit-frag.h
/usr/bin/python3 /qemu-5.2.0/meson/meson.py --internal exe --capture
ui/shader/texture-blit-frag.h -- /usr/bin/env perl
/qemu-5.2.0/scripts/shaderinclude.pl ../ui/shader/texture-blit.frag
/usr/bin/env: ‘perl’: No such file or directory
shaderinclude.pl seems to be pretty small, maybe it could be rewritten
in python?
Probably, but "make check" also requires Perl for the TAP driver. I do
have plans for using "meson test" instead, *however* there's also the
other idea I've floated of parsing the command line with Perl:
https://patchew.org/QEMU/20210107140039.467969-1-pbonz...@redhat.com/20210107140039.467969-9-pbonz...@redhat.com/
(reviews welcome by the way).
I don't like Perl really, but there's a chicken-and-egg problem between
detecting Python and using it to print the configure help script. For
configure-time tasks, Perl has the advantage that "#! /usr/bin/env perl"
just works.
- bash has to be installed explicitly. configure/meson do not check
for it, but the build will fail if they aren't present.
IIRC we were able to compile without bash before the meson conversion,
just some parts like the iotests needed the bash (at least that's why we
have a check for bash in tests/check-block.sh for example). Where is it
failing now?
It's failing due to
SHELL = /usr/bin/env bash -o pipefail
introduced in 5.2 (but not related to the Meson conversion)
commit 3bf4583580ab705de1beff6222e934239c3a0356
Author: Paolo Bonzini <pbonz...@redhat.com>
Date: Wed Oct 14 07:35:13 2020 -0400
make: run shell with pipefail
Without pipefail, it is possible to miss failures if the recipes
include pipes.
Reviewed-by: Daniel P. Berrangé <berra...@redhat.com>
Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
Paolo
