This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH - darwin] PR target/25376


Josh Conner wrote:
> Daniel Jacobowitz wrote:
>> On Mon, Feb 06, 2006 at 10:12:30AM -0800, Josh Conner wrote:
>>> I see your point, although I can't say that I agree with you.  And, I
>>> don't know of any reason why either approach (having the compiler supply
>>> a default segment name or changing the section attribute syntax) would
>>> not be possible.  Unless I'm misunderstanding the issues here, it seems
>>> to me that either change is beyond the scope of my patch, which is
>>> intending to restore support for named function sections in a manner
>>> that is consistent with named variable sections, and that appears to
>>> have been unintentionally removed.
>> I was responding specifically to the testsuite change; I don't think
>> that Darwin supports named sections from the POV of the testsuite
>> infrastructure.  For instance, look at the tests which actually
>> use dg-require-named-sections.  Looks to me like there are five of
>> them, one of which is specifically for -freorder-blocks-and-partition,
>> and the others all use section("name").  And would fail.
> 
> Ah - okay, I think I understand better where you're coming from.
> Actually, from a testsuite perspective my patch has no regressions, at
> least when I tested on darwin and arm-none-elf.  In fact, with a single
> exception (gcc.c-torture/compile/pr23237.c), all of the tests that use a
> named section attribute went from UNSUPPORTED to PASS on darwin because
> they weren't attempting assembly.  The only test that required fixing
> was pr23237.c.  But since the PR was for a compile-time failure,
> changing the test to not attempt assembly seemed like a reasonable fix
> (and was included in my patch).  Does this help alleviate your concerns?

Are there any remaining concerns that I can address on this patch?  Is
there a desire to add a distinction to the testsuite for platforms that
don't support a single-named section ((section("foo"))), even though no
tests currently require it?

Thanks -

Josh


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]