* [PATCH]: coff/pe: fix handling of weak defined symbols for --gc-sections
@ 2017-03-31 13:36 Tristan Gingold
2017-04-06 15:49 ` Nick Clifton
0 siblings, 1 reply; 3+ messages in thread
From: Tristan Gingold @ 2017-03-31 13:36 UTC (permalink / raw)
To: binutils
Hello,
this patch handles the coff/pe specific way to represent a potential
definition for a weak symbol in gc-section.
There is a link in the undefined weak symbol to the definition. So the
section for the definition must be keep if there is a reference to the
undefined symbol.
No regression on i686-pc-mingw32
Ok to commit ?
Regards,
Tristan.
diff --git a/bfd/ChangeLog b/bfd/ChangeLog
index bb851c7..e2c0103 100644
--- a/bfd/ChangeLog
+++ b/bfd/ChangeLog
@@ -1,3 +1,9 @@
+2017-03-31 gingold <gingold@gingold-Precision-7510>
+
+ * coffgen.c (_bfd_coff_gc_mark_hook): Handle PE weak
+ external symbols with a definition.
+ (_bfd_coff_gc_mark_extra_sections): Fix typo.
+
2017-03-30 Pip Cet <pipcet@gmail.com>
* elf32-wasm32.c: Add relocation code, two relocs.
diff --git a/bfd/coffgen.c b/bfd/coffgen.c
index 3c60ed4..18f423b 100644
--- a/bfd/coffgen.c
+++ b/bfd/coffgen.c
@@ -2785,8 +2785,22 @@ _bfd_coff_gc_mark_hook (asection *sec,
case bfd_link_hash_common:
return h->root.u.c.p->section;
- case bfd_link_hash_undefined:
case bfd_link_hash_undefweak:
+ if (h->symbol_class == C_NT_WEAK && h->numaux == 1)
+ {
+ /* PE weak externals. A weak symbol may include an auxiliary
+ record indicating that if the weak symbol is not resolved,
+ another external symbol is used instead. */
+ struct coff_link_hash_entry *h2 =
+ h->auxbfd->tdata.coff_obj_data->sym_hashes[
+ h->aux->x_sym.x_tagndx.l];
+
+ if (h2 && h2->root.type != bfd_link_hash_undefined)
+ return h2->root.u.def.section;
+ }
+ break;
+
+ case bfd_link_hash_undefined:
default:
break;
}
@@ -2897,7 +2911,7 @@ _bfd_coff_gc_mark_extra_sections (struct
bfd_link_info *info,
asection *isec;
bfd_boolean some_kept;
- if (bfd_get_flavour (ibfd) != bfd_target_elf_flavour)
+ if (bfd_get_flavour (ibfd) != bfd_target_coff_flavour)
continue;
/* Ensure all linker created sections are kept, and see whether
diff --git a/ld/ChangeLog b/ld/ChangeLog
index 84ef78e..e0fd169 100644
--- a/ld/ChangeLog
+++ b/ld/ChangeLog
@@ -1,3 +1,9 @@
+2017-03-31 gingold <gingold@gingold-Precision-7510>
+
+ * testsuite/ld-pe/pe.exp: New test.
+ * testsuite/ld-pe/weakdef-1.s: New test source.
+ * testsuite/ld-pe/weakdef-1.d: New test.
+
2017-03-28 Hans-Peter Nilsson <hp@axis.com>
PR ld/16044
diff --git a/ld/testsuite/ld-pe/pe.exp b/ld/testsuite/ld-pe/pe.exp
index 8d20056..b69f36a 100644
--- a/ld/testsuite/ld-pe/pe.exp
+++ b/ld/testsuite/ld-pe/pe.exp
@@ -76,6 +76,8 @@ run_dump_test "longsecn-5"
run_dump_test "orphan"
run_dump_test "orphan_nu"
+run_dump_test "weakdef-1"
+
run_dump_test "pr19803"
set pr19803_dll {
{ "PR 19803: not exporting swept symbols"
diff --git a/ld/testsuite/ld-pe/weakdef-1.d b/ld/testsuite/ld-pe/weakdef-1.d
new file mode 100644
index 0000000..57a8565
--- /dev/null
+++ b/ld/testsuite/ld-pe/weakdef-1.d
@@ -0,0 +1,8 @@
+#source: weakdef-1.s
+#target: i*86-*-cygwin* i*86-*-pe i*86-*-mingw*
+#ld: -e _start --gc-sections
+#objdump: -d
+
+#...
+ 401003: a1 00 20 40 00 mov 0x402000,%eax
+#pass
diff --git a/ld/testsuite/ld-pe/weakdef-1.s b/ld/testsuite/ld-pe/weakdef-1.s
new file mode 100644
index 0000000..ddce4fb
--- /dev/null
+++ b/ld/testsuite/ld-pe/weakdef-1.s
@@ -0,0 +1,19 @@
+ .weak _wsym
+ .section .data$wsym,"w"
+ .align 4
+_wsym:
+ .long 1
+
+ .section .text$start,"x"
+ .globl _start
+ .def _start; .scl 2; .type 32; .endef
+_start:
+ pushl %ebp
+ movl %esp, %ebp
+ movl _wsym, %eax
+ testl %eax, %eax
+ sete %al
+ movzbl %al, %eax
+ nop
+ popl %ebp
+ ret
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH]: coff/pe: fix handling of weak defined symbols for --gc-sections
2017-03-31 13:36 [PATCH]: coff/pe: fix handling of weak defined symbols for --gc-sections Tristan Gingold
@ 2017-04-06 15:49 ` Nick Clifton
2017-04-07 8:04 ` Tristan Gingold
0 siblings, 1 reply; 3+ messages in thread
From: Nick Clifton @ 2017-04-06 15:49 UTC (permalink / raw)
To: Tristan Gingold, binutils
Hi Tristan,
> Ok to commit ?
Approved - please apply.
> diff --git a/bfd/ChangeLog b/bfd/ChangeLog
> index bb851c7..e2c0103 100644
> --- a/bfd/ChangeLog
> +++ b/bfd/ChangeLog
> @@ -1,3 +1,9 @@
> +2017-03-31 gingold <gingold@gingold-Precision-7510>
> +
> + * coffgen.c (_bfd_coff_gc_mark_hook): Handle PE weak
> + external symbols with a definition.
> + (_bfd_coff_gc_mark_extra_sections): Fix typo.
> +
Cheers
Nick
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH]: coff/pe: fix handling of weak defined symbols for --gc-sections
2017-04-06 15:49 ` Nick Clifton
@ 2017-04-07 8:04 ` Tristan Gingold
0 siblings, 0 replies; 3+ messages in thread
From: Tristan Gingold @ 2017-04-07 8:04 UTC (permalink / raw)
To: Nick Clifton, binutils
On 06/04/2017 17:49, Nick Clifton wrote:
> Hi Tristan,
>
>> Ok to commit ?
>
> Approved - please apply.
Thanks, pushed.
Tristan.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2017-04-07 8:04 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-03-31 13:36 [PATCH]: coff/pe: fix handling of weak defined symbols for --gc-sections Tristan Gingold
2017-04-06 15:49 ` Nick Clifton
2017-04-07 8:04 ` Tristan Gingold
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).