[Patch] Fix for Martin's issue

[Paolo Carlini]

Hi,

so here it is, quite easy to fix, as I imagined. I was a little bit
worried that performance could be affected, so I ran three times
the performance testsuite (p4-2400, -O2) and ifstream_extract_int
is:

3.4
---
612r  600u    7s        0mem    8pf
615r  607u    6s        0mem    8pf
610r  599u    7s        0mem    8pf

3.4 + patch
-----------
619r  601u    8s        0mem    7pf
615r  601u   11s        0mem    7pf
618r  601u   13s        0mem    7pf

so, the differences are in the noise, it seems...

Otherwise, the fix is straightforard: just reorder properly the
various branches.

Tested x86-linux, will wait 'til this evening italian time.

Paolo.

///////////
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: CL_martin
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20031219/0915d9ea/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch_martin
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20031219/0915d9ea/attachment-0001.ksh>