Re: GNAT, executable stack and trampolines
On 05/24/2011 02:06 PM, Ludovic Brenta wrote:
> Reto Buerki wrote:
>> On 05/24/2011 11:00 AM, Pascal Obry wrote:
>>> ... but as you found out some work as been done to remove most cases
>>> where a stack must be executable.
>>
>> Yes. But as it seems this work has not (yet) been merged into FSF gcc.
>>
>> I'm looking into FSF gcc trunk right now to find out if the following
>> work found in the GNAT GPL 2010 compiler (file: src/gcc-43.dif) has been
>> merged:
>>
>> 2008-05-25 Eric Botcazou <ebotcazou@adacore.com>
>> Elimination of stack trampolines
>> ...[snip]...
>>
>> Part of this huge patch introduces the -fno-trampolines compiler option:
>>
>> "-fno-trampolines is enabled by default to let the compiler avoid
>> generating them if it computes that this is safe, on a case by case
>> basis, and replace them with descriptors. Descriptors are always made
>> up of data only, but the generated code must be prepared to deal with
>> them."
>>
>> This option does not exist in gcc (Debian 4.4.5-8) 4.4.5:
>> gnat1: error: unrecognized command line option "-fno-trampolines"
>>
>> It works (as expected) with GNAT GPL 2010. If you specify -ftrampolines,
>> the stack is set executable as well.
>>
>> @xavier: Can you test if such an option exists in gcc 4.6.0?
>
> I don't see this patch in
> http://gcc.gnu.org/ml/gcc-patches/2008-05/authors.html or in the ChangeLog
> in the FSF tree. The option -fno-trampolines does not appear in the index
> of the GNAT User's Guide for GCC 4.6.0. Therefore it seems like someone at
> AdaCore decided not to (or forgot to) merge this patch into the FSF tree :(
>
> I'll try to find out whether any other patches are missing. This might
> take me several days though. If anyone wants to help with this search,
> could you compare the gcc/ada/ChangeLog-2008 of GNAT GPL 2010 with that of
> GCC 4.6.0 and send me the diff? Same for all other gcc/ada/ChangeLog-*
> files, for that matter.
I can't find the file gcc/ada/ChangeLog-2008 in neither GNAT GPL 2010
sources nor binary distribution tarballs. Am I missing something?
Reply to: