Long-standing error in exception handling by pool_allocator (contains untested patch)

[Paolo Carlini]

Paolo Carlini wrote:

> Therefore, my plan would be preparing a simple testcase from your 
> complete
> analysis, regression test the slighlty tweaked version, and repost it 
> to the list
> for final review.

The below is what I have just finished regtesting on x86/x86_64-linux. 
As you
can see, I simply test that the fall back machinery originally designed 
by SGI/HP
is actually working, instead of testing for memory corruption: the 
latter is more
tricky, depends on too many parameters and risk to convey the misleading
impression that the patch is "only" about memory corruption: in that 
case simply
zeroing _S_start_free/_S_end_free before calling ::operator new would 
suffice.

I'll wait until tomorrow in case of comments...

Paolo.

////////////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CL_joachim
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20041010/ec8efac3/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_joachim
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20041010/ec8efac3/attachment-0001.ksh>