Re: [Help] New version of htslib does not build
Hi,
Am Fri, Sep 30, 2022 at 01:10:27PM +0000 schrieb John Marshall:
> Andreas Tille wrote:
> > Interestingly the missing symbols are mentioned in some warnings higher up in the changelog!
>
> As per the build log, those warnings were:
>
> Selecting previously unselected package libhtscodecs2:amd64.
> Preparing to unpack .../44-libhtscodecs2_1.2.2-1_amd64.deb ...
> Unpacking libhtscodecs2:amd64 (1.2.2-1) ...
> […]
> cram/cram_io.c:1725:23: warning: implicit declaration of function 'tok3_decode_names'; did you mean 'decode_names'? [-Wimplicit-function-declaration]
> 1725 | uint8_t *cp = tok3_decode_names(b->data, b->comp_size, &out_len);
> | ^~~~~~~~~~~~~~~~~
> | decode_names
> […]
>
> These function names (tok3_decode_names() et al) appeared in htscodecs 1.3.0 in response to [1]. Probably the underlying problem here is that you need to update the requirement on the external libhtscodecs to ensure htslib 1.16 is built against htscodecs 1.3.0.
This should be the case, thought.
> Perhaps separately, what is the reason for [2]?
>
> Commit 2905146ffe547dd98617e32d8b834f4c802a7f77
> Fix build options to make sure symbols will remain in library
>
> It would appear from the resulting additions to debian/libhts3.symbols that removing -fvisibility=hidden from DEB_CFLAGS_MAINT_APPEND has caused large numbers of HTSlib's non-HTSLIB_EXPORT-marked internal functions to be added to your libhts3.symbols file. Adding these to libhts3.symbols doesn't seem very correct (they are not part of the ABI), and I expect the upstream HTSlib maintainers will be extremely uninterested in any future complaints that non-exported symbols have changed or disappeared.
The log specified -fvisibility=hidden twice and thus I removed one instance and at
the same time adding -ffat-lto-objects which was due to the hint given by lintian
to fix a lintian error. Checking the build logs[3] I realise that my attempt to
remove a duplicate of -fvisibility=hidden has removed both instances - argh!
I think this should be re-added - thanks a lot for watching me!
Strangely enough the autopkgtest is failing[4] (which was not the case in my
local build.) Could anybody have a look?
Thanks again for your valuable contributions
Andreas.
> [1] https://github.com/samtools/htscodecs/issues/36
> [2] https://salsa.debian.org/med-team/htslib/-/commit/2905146ffe547dd98617e32d8b834f4c802a7f77
[3] https://salsa.debian.org/med-team/htslib/-/jobs/3320177
[4] https://salsa.debian.org/med-team/htslib/-/jobs/3320185
--
http://fam-tille.de
Reply to: