[Bug binutils/27666] ar doesn't create indices on Solaris/sparcv9

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://sourceware.org/bugzilla/show_bug.cgi?id=27666

--- Comment #3 from Rainer Orth <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #2 from Nick Clifton <nickc at redhat dot com> ---
> Created attachment 13455
>   --> https://sourceware.org/bugzilla/attachment.cgi?id=13455&action=edit
> Proposed patch

Hi Nick,

sorry for the long delay: I've been on vacation for some time.

>   Would you mind trying out the uploaded patch ?
>
>   I am working on a theory that the real bug is that
> _bfd_write_archive_contents() is not creating the symbol index.  As you noted
> the call to bfd_check_format() fails, because there are two potential 
> matches. 
> But I do not think that this is grounds for the archiver to decide that there
> are no objects in the library.
>
>   Even with this patch applied, most of the linker tests that are failing
> continue to do so, but I have not yet investigated why.  I just wanted to show
> you my ida, and see if you thought that it might be worth persuing.

As a first step, I've tested the patch on both amd64-pc-solaris2.11 (no
changes in test results) and sparcv9-sun-solaris2.11.  The situation is
mixed here:

@@ -45 +44,0 @@
-FAIL: --export-dynamic-symbol foo archive
@@ -48 +46,0 @@
-FAIL: --export-dynamic-symbol-list foo archive
@@ -55 +53,2 @@
-FAIL: ld link shared library
+FAIL: ld export symbols from archive

spawn [open ...]^M
0000000000100250 A _DYNAMIC
0000000000100330 A _GLOBAL_OFFSET_TABLE_
0000000000100400 A _PROCEDURE_LINKAGE_TABLE_
                 U exclude_sym
0000000000100338 D include_sym

+FAIL: ld don't exclude symbols from archive - --exclude-libs foo:bar

spawn [open ...]^M
0000000000100250 A _DYNAMIC
0000000000100330 A _GLOBAL_OFFSET_TABLE_
0000000000100400 A _PROCEDURE_LINKAGE_TABLE_
                 U exclude_sym
0000000000100338 D include_sym

@@ -109 +107,0 @@
-FAIL: ld-misc/defsym1

In both failing cases, exclude_common is missing.  The symbol is present
in exclude2.o, but missing from exclude.so.

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