[Bug binutils/28719] DWARF-5 section names in PE/PEP and weak symbols in Cygwin

[nickc at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28719

--- Comment #7 from Nick Clifton <nickc at redhat dot com> ---
(In reply to Achim from comment #6)
> So it turns out that removing the section at that commit (as I had done
> during bisecting) would actually restore the previous behaviour (the
> executable would not run).  Applying the same fix on top of both the 2.37
> and 2.38 releases doesn't, however.  So it seems that one of the later
> commits would be responsible for the actual resolution of the weak symbol.

Hmm, I suspect that this is going to be hard to track down.

One thing I am not sure about however is the involvement of the cygwin.a
library.
The trace you reported in comment #4 shows that it is being used to resolve the
weak reference to fputs.  Presumably this should not be happening, but why not
?  Should cygwin.a not be included in the link ?  Or should it be included but
not used to resolve weak references ?  Or ... included, used to resolve
non-weak references, but if a member of the archive is used to resolve a
non-weak reference then that member can then also be used to resolve weak
references.  In which case, the problem is why is the t-d000578.o member of
cygwin.a being pulled in in the first place ?

(I am also wondering if this problem has anything to do with the changes made
to the xcoff_link_check_ar_symbols() function in bfd/xcofflink.c).

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