[PATCH] Unbound alloca in crypt routines

[Carlos O'Donell]

On Thu, Mar 29, 2012 at 1:22 PM, Jeff Law <law@redhat.com> wrote:
> On 03/29/2012 11:10 AM, Roland McGrath wrote:
>>
>> The pontificating comments are inappropriate.  The use of
>> __libc_use_alloca is well established throughout the code.
>> It does not require any special comment, let alone a snarky one.
>
> Just because it's well established throughout the code doesn't mean using
> alloca is good practice.  In fact, I would argue that the number of fixes
> folks have had to make to deal with unbound allocas is good evidence that
> the practice of using alloca is bad.

The discussion of alloca is a separate conversation. If you want to
have that conversation we certainly can. Please start another thread
and we can take it from there.

> However, I'll certainly remove that part of the comment if you object to it,
> and it sounds like you do.  Aside from that part of the comment, what others
> issues need to be addressed?

I just finished reviewing the patch, and technically it looks good to
me, modulo Marek's comments.

What kind of testing did you do here?

Do you have a test case we could add to the testsuite that exercises
the new malloc case?

I would like to try get a testcase added for fixes like this to
prevent them from regressing.

Cheers,
Carlos.