[Bug debugedit/33823] debugedit fails to detect sources if DWARF contains relative, not absolute paths (empty debugsourcefiles.list)

[david.abdurachmanov at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=33823

--- Comment #10 from David Abdurachmanov <david.abdurachmanov at gmail dot com> ---
(In reply to Mark Wielaard from comment #9)
> (In reply to David Abdurachmanov from comment #8)
> > Shouldn't we make debugedit fail in this case?
> 
> But how do we know it "failed"?
> You asked it for a list of files starting with a certain prefix in the
> (debug) file.
> It didn't find any with that prefix. Is that a failure or are there just not
> any such files?

If the list produced by debugedit is empty or/and all provided paths in output
are empty directories? Not sure about directory thing, but that's why it
doesn't fail in the main Koji. We simply package empty directories.

You can also detect flags being used if compiler recorded them.
Annobin/Annocheck does that.

For example, random annocheck scratch I was doing to solve other problem:
https://riscv-kojipkgs.fedoraproject.org//work/tasks/2186/122186/build.log

I wonder if annobin records these two.

> 
> > We are failing on riscv64 because of a different issue. On non-riscv64 arch
> > (e.g. x86_64, aarch64, etc.) you get a directory and it's get packaged. On
> > riscv64 we don't get directory reported by debugedit and thus
> > debugsourcefiles.list is empty causing RPM to fail it.
> 
> Which directory/path is this?

https://koji.fedoraproject.org/koji/rpminfo?rpmID=46280876

It doesn't fail in main Koji because on all other arches you get:
/usr/src/debug/opensurge-0.6.1.3-1.fc44.aarch64
/usr/src/debug/opensurge-0.6.1.3-1.fc44.aarch64/redhat-linux-build

On riscv64 debugedit returns empty list (no directories listed). That's why we
fail on riscv64. So there is this difference to understand too.

>  
> > If debugedit cannot handle it should fail with an error.
> 
> But we have to figure out what "it" is.
> This bug is about debugedit not handling "relative" paths (as if you used
> -fdebug-prefix-map=$(pwd)=.
> But "this" seems to be something else.
> 
> > Again, we also need to debug why debugedit provides different output on
> > riscv64.
> 
> Do you happen to have a x86_64 and riscv64 file that show this?

Not at this point. I would need to spin a build. The packages are mentioned in
the ticket.

-- 
You are receiving this mail because:
You are on the CC list for the bug.