[Bug gdb/29999] Bogus error message printed when debuginfod doesn't find source file

["amerey at redhat dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29999

--- Comment #1 from Aaron Merey <amerey at redhat dot com> ---
Created attachment 14592
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14592&action=edit
patch

(In reply to Simon Marchi from comment #0)
> I think this use of errno, in print_source_lines_base, is incorrect:
> 
> https://gitlab.com/gnutools/binutils-gdb/-/blob/
> 6f9f448118eaeaf006f867a25699aef7d8c72770/gdb/source.c#L1357
> 
> The last thing we did before this was to call into debuginfod_find_source,
> which returned -2 (meaning ENOENT, which makes sense).  But the errno value
> after calling debuginfod_find_source is not defined, and I think this
> results in an error message that does not make sense.

It looks like errno gets set in find_and_open_source (in the case of ENOENT at
least) before the call to debuginfod_find_source:

https://gitlab.com/gnutools/binutils-gdb/-/blob/6f9f448118eaeaf006f867a25699aef7d8c72770/gdb/source.c#L1179

However function calls within debuginfod_find_source may end up reassigning
errno to a value unrelated to the status of the download, even if the download
is successful. debuginfod_find_source returns a negative errno indicating the
actual reason for a failed download. This is the errno that should be reported
to the user, not one related to routine cache creation, for example.

We also don't want to hide an errno set in find_and_open_source. debuginfod
errors are already reported in the debuginfod progress output, so I think we
should save errno before calling debuginfod_find_source and revert errno to the
saved value afterwards. This way both debuginfod errors and
find_and_open_source errors are correctly reported to the user.

I've attached a patch for this.

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