Summary: | [4.5 regression] RTL checking failure compiling libgcc | ||
---|---|---|---|
Product: | gcc | Reporter: | Eric Botcazou <ebotcazou> |
Component: | debug | Assignee: | Jakub Jelinek <jakub> |
Status: | RESOLVED FIXED | ||
Severity: | critical | CC: | aoliva, gcc-bugs |
Priority: | P3 | ||
Version: | 4.5.0 | ||
Target Milestone: | 4.5.0 | ||
Host: | ix86-*-linux | Target: | ix86-*-linux |
Build: | ix86-*-linux | Known to work: | |
Known to fail: | Last reconfirmed: | ||
Attachments: |
Reduced testcase.
gcc45-pr41236.patch |
Description
Eric Botcazou
2009-09-03 08:07:35 UTC
Created attachment 18470 [details]
Reduced testcase.
To be compiled at -O -g with an --enable-checking=yes,rtl compiler.
Created attachment 18474 [details] gcc45-pr41236.patch Either this, or just using XEXP (rtl, 0) instead of SUBREG_REG (rtl). This is only RTL checking problem, without RTL checking it will work just fine. Doing full --enable-checking=yes,rtl bootstrap now. That still assumes that inside of SIGN_EXTEND/ZERO_EXTEND is a REG, which isn't always the case. Testing with loc_descriptor call there instead now. Subject: Bug 41236 Author: jakub Date: Thu Sep 3 16:33:27 2009 New Revision: 151385 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151385 Log: PR debug/41236 * dwarf2out.c (loc_descriptor): Don't use SUBREG_REG macro on SIGN_EXTEND or ZERO_EXTEND. Don't assume there is a REG inside of it or SUBREG. Modified: trunk/gcc/ChangeLog trunk/gcc/dwarf2out.c Fixed. |