Re: [Valgrind-developers] Valgrind-3.22.0.RC2 is available for testing

[Mark W. <ma...@kl...>]

Hi Matthias,

On Sat, 2023-10-28 at 11:23 +0200, zz...@ge... wrote:
> Am 26.10.23 um 14:48 schrieb Mark Wielaard:
> > An RC2 tarball for 3.22.0 is now available at
> > https://sourceware.org/pub/valgrind/valgrind-3.22.0.RC2.tar.bz2
> > (md5sum = 07bb18b0fd1c7347b350d8d71dc940f6)
> > (sha1sum = 61c29e47efdc933ea3f23cb2f0fcebcd6d96dab1)
> > https://sourceware.org/pub/valgrind/valgrind-3.22.0.RC2.tar.bz2.asc
> > 
> 
> After fixing a ccache issue and looking into various things I found 
> these minor issues:

Thanks for doing this. What was the ccache issue?

> Comparing the distributed tarball to a locally generated one (make dist) 
> I found these differences:
> 
> FAQ.txt:
> -Release 3.21.0 28 Apr 2023
> +Release 3.22.0.RC2 26 Oct 2023

I see we seem to generate the FAQ.txt in the top-level dir, but then
distribute the one in docs/FAQ.txt.

In docs/Makefile.am we have:

# This is done at 'make dist' time.  It builds the html docs, print
# docs and man pages and copies them into the docs/ directory in the
# tarball.
 ifeq ($(BUILD_ALL_DOCS),yes)
dist-hook: FAQ.txt html-docs man-pages print-docs
        cp -r html $(distdir)
        cp FAQ.txt $(distdir)/..
        cp *.1 $(distdir)
        cp print/index.pdf $(distdir)
        cp print/index.ps $(distdir)
 else
dist-hook: FAQ.txt html-docs man-pages
        cp -r html $(distdir)
        cp FAQ.txt $(distdir)/..
        cp *.1 $(distdir)
 endif

But in the top-level Makefile.am we don't seem to reference the
FAQ.txt.

How odd. I don't immediately know what was intended here.

> A lot of filter-scripts change:
> -SED=/usr/bin/sed
> +SED=/bin/sed
> 
>   .../drd/tests/filter_error_count        | 2 +-
>   .../drd/tests/filter_error_summary      | 2 +-
>   .../drd/tests/filter_stderr             | 2 +-
>   ...lter_stderr_and_thread_no_and_offset | 2 +-
>   .../drd/tests/filter_thread_no          | 2 +-
>   .../drd/tests/filter_xml_and_thread_no  | 2 +-
>   .../gdbserver_tests/filter_gdb          | 2 +-
>   .../filter_memcheck_monitor             | 2 +-
>   .../gdbserver_tests/filter_stderr       | 2 +-
>   .../gdbserver_tests/filter_vgdb         | 2 +-
>   .../helgrind/tests/filter_stderr        | 2 +-
>   .../memcheck/tests/filter_dw4           | 2 +-
>   .../memcheck/tests/filter_overlaperror  | 2 +-
>   .../memcheck/tests/filter_stderr        | 2 +-
>   .../memcheck/tests/x86/filter_pushfpopf | 2 +-
>   .../tests/filter_discards               | 2 +-
>   .../tests/filter_stderr_basic           | 2 +-
> 
> 
> It should not break anything as the filter-scripts are re-generated 
> while building. But I wonder why they are in the tarball at all.
> 
> see memcheck/tests/Makefile.am:
> dist_noinst_SCRIPTS = \
> [...]
>          filter_stderr filter_xml \

You are right, all those filter scripts generated by configure.ac by
AC_CONFIG_FILES from the .in files should not be in dist_noinst_SCRIPTS
but in noinst_SCRIPTS.

I am testing a patch to do that, but might want to postpone that fix
till after the final release because I am a little afraid to mess
something up and it is a minor issue.

Thanks,

Mark