* [Patch gas/mach-o] try to specify section flags better
@ 2012-02-18 11:51 Iain Sandoe
2012-02-20 16:37 ` Tristan Gingold
0 siblings, 1 reply; 3+ messages in thread
From: Iain Sandoe @ 2012-02-18 11:51 UTC (permalink / raw)
To: binutils Development; +Cc: Tristan Gingold
I found in some tests, that:
align x
was causing the alignment padding to be zeroes - and thus broken.
This was caused by section definitions from GCC for various hot/cold/
startup etc. sections which are specified the long-hand way (rather
than via a canonical name).
The solution is to examine the incoming information and try to set
the section flags from that, if possible. This is adequate for the
GCC output - but, in the general case it might be wise for people to
write
align x,<nop>
A similar scenario causes relocation problems if we don't recognize
debug sections.
OK?
Iain
gas:
* config/obj-macho.c (obj_mach_o_make_or_get_sect): In the absence of
canonical information, try to determine CODE and DEBUG section flags
from the mach-o section data.
--- gas/config/obj-macho.c 2012-02-12 15:42:47.000000000 +0000
+++ gas/config/obj-macho-b.c 2012-02-18 11:42:43.000000000 +0000
@@ -240,6 +240,16 @@ obj_mach_o_make_or_get_sect (char * segn
if (oldflags == SEC_NO_FLAGS)
{
+ if (flags == SEC_NO_FLAGS
+ && (specified_mask & SECT_ATTR_SPECIFIED)
+ && (secattr & BFD_MACH_O_S_ATTR_PURE_INSTRUCTIONS))
+ flags |= SEC_CODE;
+
+ if (flags == SEC_NO_FLAGS
+ && (specified_mask & SECT_ATTR_SPECIFIED)
+ && (secattr & BFD_MACH_O_S_ATTR_DEBUG))
+ flags |= SEC_DEBUGGING;
+
/* New, so just use the defaults or what's specified. */
if (! bfd_set_section_flags (stdoutput, sec, flags))
as_warn (_("failed to set flags for \"%s\": %s"),
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Patch gas/mach-o] try to specify section flags better
2012-02-18 11:51 [Patch gas/mach-o] try to specify section flags better Iain Sandoe
@ 2012-02-20 16:37 ` Tristan Gingold
2012-02-20 19:45 ` Iain Sandoe
0 siblings, 1 reply; 3+ messages in thread
From: Tristan Gingold @ 2012-02-20 16:37 UTC (permalink / raw)
To: Iain Sandoe; +Cc: binutils Development
On Feb 18, 2012, at 12:50 PM, Iain Sandoe wrote:
> I found in some tests, that:
> align x
> was causing the alignment padding to be zeroes - and thus broken.
>
> This was caused by section definitions from GCC for various hot/cold/startup etc. sections which are specified the long-hand way (rather than via a canonical name).
>
> The solution is to examine the incoming information and try to set the section flags from that, if possible. This is adequate for the GCC output - but, in the general case it might be wise for people to write
> align x,<nop>
>
> A similar scenario causes relocation problems if we don't recognize debug sections.
>
> OK?
Yes. That makes sense.
I think you should add a comment here too (just the one from ChangeLog would be ok).
Tristan.
> Iain
>
>
> gas:
> * config/obj-macho.c (obj_mach_o_make_or_get_sect): In the absence of
> canonical information, try to determine CODE and DEBUG section flags
> from the mach-o section data.
>
> --- gas/config/obj-macho.c 2012-02-12 15:42:47.000000000 +0000
> +++ gas/config/obj-macho-b.c 2012-02-18 11:42:43.000000000 +0000
> @@ -240,6 +240,16 @@ obj_mach_o_make_or_get_sect (char * segn
>
> if (oldflags == SEC_NO_FLAGS)
> {
> + if (flags == SEC_NO_FLAGS
> + && (specified_mask & SECT_ATTR_SPECIFIED)
> + && (secattr & BFD_MACH_O_S_ATTR_PURE_INSTRUCTIONS))
> + flags |= SEC_CODE;
> +
> + if (flags == SEC_NO_FLAGS
> + && (specified_mask & SECT_ATTR_SPECIFIED)
> + && (secattr & BFD_MACH_O_S_ATTR_DEBUG))
> + flags |= SEC_DEBUGGING;
> +
> /* New, so just use the defaults or what's specified. */
> if (! bfd_set_section_flags (stdoutput, sec, flags))
> as_warn (_("failed to set flags for \"%s\": %s"),
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Patch gas/mach-o] try to specify section flags better
2012-02-20 16:37 ` Tristan Gingold
@ 2012-02-20 19:45 ` Iain Sandoe
0 siblings, 0 replies; 3+ messages in thread
From: Iain Sandoe @ 2012-02-20 19:45 UTC (permalink / raw)
To: Tristan Gingold; +Cc: binutils Development
On 20 Feb 2012, at 16:37, Tristan Gingold wrote:
>> OK?
>
> Yes. That makes sense.
> I think you should add a comment here too (just the one from
> ChangeLog would be ok).
thanks, done & applied.
Iain
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2012-02-20 19:45 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-02-18 11:51 [Patch gas/mach-o] try to specify section flags better Iain Sandoe
2012-02-20 16:37 ` Tristan Gingold
2012-02-20 19:45 ` Iain Sandoe
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).