[Pkg-phototools-devel] Bug#989222: Bug#989222: Bug#989222: darktable: crashes with "Floating point exception (core dumped)" after loading some DNG files

[Jonas Smedegaard]

Quoting David Bremner (2021-05-29 18:58:16)
> Jonas Smedegaard <dr at jones.dk> writes:
> > Quoting David Bremner (2021-05-29 14:32:09)
> >> Jonas Smedegaard <dr at jones.dk> writes:
> >> > Upstream tracked and has solved this issue at 
> >> > https://github.com/darktable-org/darktable/issues/8951
> >> >
> >> > At upstream issue tracking is mentioned that the file they used 
> >> > as proof was arguably broken.
> >> >
> >> 
> >> Upstream suggests that fix is not easily backportable,
> >
> > Oh, wauw - then apparently I did something hard without even knowing 
> > it: I grabbed and applied the patch and it seems to work just fine: 
> > https://source.puri.sm/pureos/packages/darktable/-/commit/19eb57c7bffdca0575f1d526656ca4614f05a33c
> >
> > Where do upstream say that?  Perhaps they mean backporting further 
> > back than nearest point release, 3.4.1?
> >
> 
> They mentioned there was a substantial amount of code deletion.

Seems to me that upstream code changes consist of a) fixing the bug 
(involving edits to 7 lines) and b) dropping hand-optimized parallel 
implementation for SSE-capable platforms lots and lot of lines dropped), 
where b) is all ifdef'ed by __SSE__.

Seems to me that upstream talked about dropping that SSE code causing an 
acceptable 20% slowdown of that routine.

I still don't see anywhere that upstream express concern over 
backportability of the bugfix.  If you meant to say that the _size_ of 
upstream changes implied it not being easily backportable, then please 
consider look closer and see if you can agree with me that essentially 
it is a 7 lines change + a massive removal of an optimization deemed no 
longer beneficial to maintain upstream.


> Also, there were 4 commits in that PR; maybe some are harder to 
> backport than others.

Seems to me that the 4 patches in the PR are equivalent to git commit 
2ff4fc58e442cec464444042c0be4ac84352196b in upstream master branch, 
which is what I chery-picked and tested.


Thanks for taking time to look at this,

 - Jonas

-- 
 * Jonas Smedegaard - idealist & Internet-arkitekt
 * Tlf.: +45 40843136  Website: http://dr.jones.dk/

 [x] quote me freely  [ ] ask before reusing  [ ] keep private
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-phototools-devel/attachments/20210529/7bb74d41/attachment.sig>