[3.3] Fix PR target/11689 (x86)

[Eric Botcazou]

Hi,

This is yet another testcase for which GCC produces unassemblable code for 
the K6 family of processors, a regression from GCC 3.2.3 present on the 3.3 
branch.

As usual, the compiler produces out-of-range 'loop' insns because of 
underestimated instruction lengths. The bug has been latent since egcs.

Bootstrapped/regtested on k6-redhat-linux-gnu (c,c++,objc,f77 3.3 branch).
Ok for 3.3 branch and mainline?


2003-09-06  Eric Botcazou  <ebotcazou@libertysurf.fr>

        PR target/11689
        * config/i386/i386.c (memory_address_length): Fix computation when
	the base is esp or ebp.

2003-09-06  Eric Botcazou  <ebotcazou@libertysurf.fr>

        * g++.dg/opt/longbranch2.C: New test.


-- 
Eric Botcazou
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pr11689.diff
Type: text/x-diff
Size: 1394 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20030906/939c4529/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: longbranch2.C
Type: text/x-c++src
Size: 1684 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20030906/939c4529/attachment-0001.bin>