Re: dwz 0.14 release?

[Mark Wielaard <mark@klomp.org>]

Hi Tom,

On Fri, Feb 12, 2021 at 11:08:42AM +0100, Mark Wielaard wrote:
> > > I haven't really investigated why that is. But we can always say
> > > that ODR support is experimental and doesn't yet work for DWARF5.
> > 
> > I can't reproduce this, can you open an PR with more details?
> 
> https://sourceware.org/bugzilla/show_bug.cgi?id=27400
> Let me know if you need any test binaries and I'll attach them to the
> bug.

So, this was fixed already. And we found some subtle other bugs.

> > Anyway, odr will be experimental.  It still need to marked as such.
> 
> Also opened a bug for that:
> https://sourceware.org/bugzilla/show_bug.cgi?id=27401
> 
> And another to document the status of DWARF 5:
> https://sourceware.org/bugzilla/show_bug.cgi?id=27402
> 
> I'll resolve that by updating dwz.1 with an overview of the current
> support for DWARF 5 in dwz.

I just pushed a patch for this.

> > Furthermore, I still need to go through the PR list and clean up.
> 
> We have about 50 open bugs. I can go through them and see if any of
> them is a showstopper/regression since 0.13, but given that various
> distros have switched to current git trunk already I think what we have
> now is consistently better than 0.13. What would be the most convenient
> to mark up the bugs?

There are 53 open bugs right now. I looked briefly at all of
them. Most are ideas for improvements of various forms. For more
optimizations or supporting certain DWARF constructs. A couple simply
don't have enough information to replicate the issue. Or are for
things that are hard to support, like the mips bug (26738) or the
emit-relocs issue (24345).

I don't believe any of the bugs are regressions since 0.13. But the
following 6 bugs seem good to take a quick look at to see if a fix is
easy, but none of them seem like real showstoppers for a
release. Since there are a lot of things fixed and new features since
0.13 I think we should do a release as soon as we know these aren't
fixed in a couple of days.

https://sourceware.org/bugzilla/show_bug.cgi?id=26252
[odr] dwz.c:11404: write_die: Assertion `value && refdcu->cu_kind != CU_ALT'
      failed. #2 
For which you just posted a patch.

https://sourceware.org/bugzilla/show_bug.cgi?id=24275
hardlink handling leaves temporary file if not file compressed
This is mostly an annoyance, but I haven't figured out the code to fix it.

https://sourceware.org/bugzilla/show_bug.cgi?id=27401
Document that the --odr flag is currently experimental
Would be nice to have documented what works and what still needs work
There are a couple of ODR bugs for which I didn't know the current
status, e.g. Bug 24198 (which is the tracker bug?)

https://sourceware.org/bugzilla/show_bug.cgi?id=27363
Emit more detailed diagnostic output for "Unknown DWARF" 
This has improved a bit. I don't think we need to go so far as the
reporter would like. But we could add a bit more output to some of the
errors like we did in commit 4705796eb "Add DIE offsets in error
messages to make it easier to find what is wrong."

https://sourceware.org/bugzilla/show_bug.cgi?id=25252
dwz: returns exit status 1, causing FTBFS in deal.ii
This seems partially fixed, but still has a patch that looks
plausible, but I don't really know if it is still needed.

https://sourceware.org/bugzilla/show_bug.cgi?id=25459
Forward pseudo-reference triggers error
Seems like a real issue. And has a test binary that triggers the
issue. On the other hand this test binary is slightly odd. It is
referenced in a couple of other bug reports, but it seems unclear how
it was ever generated. This might not be easily fixable without adding
a new pass over the whole DIE tree. Unless someone has a clever idea.

Cheers,

Mark