This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] New ia64 @slotcount pseudo func
>>> Douglas B Rupp <rupp@gnat.com> 29.10.09 09:31 >>>
>Jan Beulich wrote:
>>>>> Douglas B Rupp <rupp@gnat.com> 28.10.09 21:24 >>>
>>> Submitted for comment as per previous discussions.
>>
>> Why can't the debugger do the math that you do here? Unless Intel's
>> assembler is intending to also support @slotcount, I'm not certain it's a
>> good idea to add such to gas.
>
>This is for the VMS Debugger on IA64. VMS compilers generate object code
>directly and don't use the Intel assembler. The debugger requires a
>slotcount here. There's some history on the gcc list if your interested.
To me this at a first glance looks like the debugger would need fixing
instead of the assembler...
>>
>> Also:
>>
>>> + /* Calculate the number of instruction slots between the two
>>> + labels. They are guaranteed to be in the same (.text)
>>> + section. */
>>> + end = S_GET_VALUE (symbolPend);
>>> + beg = S_GET_VALUE (symbolPbeg);
>>
>> Where is this "guarantee" being enforced? I think you mean "They are
>> supposed...", and need to check this really is the case.
>
>It's only used for function prologue, epilogue offsets which are always
>going to be in the same .text section, right? If you foresee using this
>pseudo func for something else, then it needs a little more error
>checking or I can add some error checking jic.
You intend to use it only that way, which doesn't mean others couldn't
start using it for other things. Unless its use in other than the intended
context gets reported as an error, imo it should make no assumptions
about where it got used.
Btw., couldn't (if you really can't fix the debugger) the addition be made
VMS-specific (so that on platforms where the Intel assembler can be
considered the reference assembler no incompatible extensions become
available)? Alternatively, could you get Intel's buyoff on reserving the
"slotcount" pseudo function name as an optional extension with precisely
the semantics you need?
Jan