|
From: Nicholas N. <nj...@cs...> - 2008-05-27 22:28:20
|
On Tue, 27 May 2008, John Reiser wrote:
>>> On x86, what is the status of getting FPU and SSE register state
>>> saved into (and restored from) the signal context, so that a signal handler
>>> can inspect and modify the values of FPU and SSE registers?
>>
>> It looks like the comment you quoted pretty much answers your
>> question doesn't it?
>
> No, the comment does not necessarily answer, "What is the status?"
> According to "svn blame sigframe-x86-linux.c", those lines are from
> r3265, and according to "svn log sigframe-x86-linux.c", r3265 was
> added on 2005-03-10, which is more than three years ago.
>
> Comments are not analyzed by the compiler.
> Comments are not checked by the regression tests.
> Comments are not exercised by running user code through the tools.
> Thus the natural tendency is for comments to become stale,
> particularly comments that employ future tense ("will be fixed.")
> Also, in a project such as valgrind the state of svn does not
> necessarily provide a good window into the state of activity.
>
> I was fishing for news about anyone else working in this area.
> By the [lack of] responses, so far I see that nobody else is.
John, your emails on this list are typically well-informed, but they are
often also terse, cryptic and/or aggressive. If you write that way, you
shouldn't be surprised if people respond in kind.
With respect to the particular FPU/SSE state problem: if you explain why
you need this shortcoming fixed -- eg. what program(s) it will enable --
that might help motivate someone to fix it. Your uses of Valgrind tend to
be more specialised and demanding than most people's, and obscure bugs tend
to be lower priority.
Nick
|