Re: [PATCH] binutils, gdb: support zstd compressed debug sections

[Jan Beulich <jbeulich@suse.com>]

On 19.09.2022 08:55, Simon Marchi via Binutils wrote:
>> diff --git a/config/zstd.m4 b/config/zstd.m4
>> new file mode 100644
>> index 00000000000..b8b3a199730
>> --- /dev/null
>> +++ b/config/zstd.m4
>> @@ -0,0 +1,12 @@
>> +AC_DEFUN([AM_ZSTD],
>> +[
>> +  zstdlib=
>> +  AC_ARG_WITH(system-zstd,
>> +  [AS_HELP_STRING([--with-system-zstd], [use installed zstd])],
>> +  if test x$with_system_zstd = xyes ; then
>> +    zstdlib=-lzstd
>> +    AC_DEFINE(HAVE_ZSTD_H)
>> +  fi
>> +  )
>> +  AC_SUBST(zstdlib)
>> +])
> 
> Hi Fangrui,
> 
> I understand that you did this by copying zlib here.  However, the only
> reason we have a --use-system-zlib is because the binutils-gdb contains
> a copy of zlib, which is used by default, and --use-system-zlib can be
> used to prefer the system version.  Since we don't carry a copy of zstd
> in the tree, I don't think we need --with-system-zstd.
> 
> However, since zstd ships with a .pc file, my preference would be to use
> PKG_CHECK_MODULES to have pkg-config find the right flags, and have
> --with/--without-zstd.  Same as I did for the msgpack library here:
> 
>   https://gitlab.com/gnutools/binutils-gdb/-/commit/2952f10cd79af4645222f124f28c7928287d8113
> 
> I am not a binutils maintainers though (just GDB), so in the end it's up
> to the binutils maintainers to decide and GDB will probably follow suit.

FWIW I agree with the comment.

Jan