GCC Bugzilla – Bug 29132
[4.1 Regression] Mips exception handling broken.
Last modified: 2006-09-25 03:23:05 UTC
The combination of GCC (r116604 or later) and binutils-2.17 creates bad .eh_frame data causing SIGSEGV at runtime anytime an exception is thrown. To reproduce just run the g++ testsuite. All exception tests FAIL. This problem is discussed further in: http://gcc.gnu.org/ml/gcc-patches/2006-09/msg00732.html
Hi David, I was wondering if you have a MIPS tree handy, whether you could easily test the following single line patch: Index: dwarf2out.c =================================================================== *** dwarf2out.c (revision 117035) --- dwarf2out.c (working copy) *************** dwarf2out_begin_prologue (unsigned int l *** 2572,2578 **** fde = &fde_table[fde_table_in_use++]; fde->decl = current_function_decl; fde->dw_fde_begin = dup_label; ! fde->dw_fde_current_label = NULL; fde->dw_fde_hot_section_label = NULL; fde->dw_fde_hot_section_end_label = NULL; fde->dw_fde_unlikely_section_label = NULL; --- 2572,2578 ---- fde = &fde_table[fde_table_in_use++]; fde->decl = current_function_decl; fde->dw_fde_begin = dup_label; ! fde->dw_fde_current_label = dup_label; fde->dw_fde_hot_section_label = NULL; fde->dw_fde_hot_section_end_label = NULL; fde->dw_fde_unlikely_section_label = NULL; Due to all the abstraction with debugging formats, its difficult to tell the order in which things get executed, and whether this initial value for dw_fde_current_label survives long enough to avoid use of a set_loc. Many thanks in advance,
The patch also went into the 4.1 branch.
Subject: Re: [4.2 Regression] Mips exception handling broken. roger at eyesopen dot com wrote: > ------- Comment #1 from roger at eyesopen dot com 2006-09-18 21:27 ------- > Hi David, > > I was wondering if you have a MIPS tree handy, whether you could easily > test the following single line patch: > > Index: dwarf2out.c > =================================================================== > *** dwarf2out.c (revision 117035) > --- dwarf2out.c (working copy) > *************** dwarf2out_begin_prologue (unsigned int l > *** 2572,2578 **** > fde = &fde_table[fde_table_in_use++]; > fde->decl = current_function_decl; > fde->dw_fde_begin = dup_label; > ! fde->dw_fde_current_label = NULL; > fde->dw_fde_hot_section_label = NULL; > fde->dw_fde_hot_section_end_label = NULL; > fde->dw_fde_unlikely_section_label = NULL; > --- 2572,2578 ---- > fde = &fde_table[fde_table_in_use++]; > fde->decl = current_function_decl; > fde->dw_fde_begin = dup_label; > ! fde->dw_fde_current_label = dup_label; > fde->dw_fde_hot_section_label = NULL; > fde->dw_fde_hot_section_end_label = NULL; > fde->dw_fde_unlikely_section_label = NULL; > > Due to all the abstraction with debugging formats, its difficult to tell the > order in which things get executed, and whether this initial value for > dw_fde_current_label survives long enough to avoid use of a set_loc. > I will try the patch now. A full build with libgcj will take me several hours to build and test. David Daney.
Subject: Re: [4.2 Regression] Mips exception handling broken. roger at eyesopen dot com wrote: > ------- Comment #1 from roger at eyesopen dot com 2006-09-18 21:27 ------- > Hi David, > > I was wondering if you have a MIPS tree handy, whether you could easily > test the following single line patch: > > Index: dwarf2out.c > =================================================================== > *** dwarf2out.c (revision 117035) > --- dwarf2out.c (working copy) > *************** dwarf2out_begin_prologue (unsigned int l > *** 2572,2578 **** > fde = &fde_table[fde_table_in_use++]; > fde->decl = current_function_decl; > fde->dw_fde_begin = dup_label; > ! fde->dw_fde_current_label = NULL; > fde->dw_fde_hot_section_label = NULL; > fde->dw_fde_hot_section_end_label = NULL; > fde->dw_fde_unlikely_section_label = NULL; > --- 2572,2578 ---- > fde = &fde_table[fde_table_in_use++]; > fde->decl = current_function_decl; > fde->dw_fde_begin = dup_label; > ! fde->dw_fde_current_label = dup_label; > fde->dw_fde_hot_section_label = NULL; > fde->dw_fde_hot_section_end_label = NULL; > fde->dw_fde_unlikely_section_label = NULL; > > Due to all the abstraction with debugging formats, its difficult to tell the > order in which things get executed, and whether this initial value for > dw_fde_current_label survives long enough to avoid use of a set_loc. > > Many thanks in advance, > Preliminary results using: readelf --debug-dump=frames libstdc++.so show that this patch is working. I now get: 00000064 00000024 00000068 FDE cie=00000000 pc=0003ed40..0003eed4 Augmentation data: 00 00 00 00 DW_CFA_advance_loc: 16 to 0003ed50 DW_CFA_def_cfa_offset: 48 DW_CFA_advance_loc: 24 to 0003ed68 DW_CFA_offset: r16 at cfa-24 DW_CFA_offset: r17 at cfa-20 DW_CFA_offset: r18 at cfa-16 DW_CFA_offset: r19 at cfa-12 DW_CFA_offset: r20 at cfa-8 Both addresses shown for DW_CFA_advance_loc are within the range of PC for the FDE. It will take (a lot) more time to see if the code actually works on my test platform, but it looks good to me so far. David Daney.
OK, I finished running make -k check with a i686-linux -> mipsel-linux cross compiler on my mips board for c, c++, java and there are only the 'normal' failures (mostly attributable to running the testsuite on a remote board). Exception handling seems to be fixed with the Roger's patch applied. I would say test it on a couple of other targets of your choice and then commit it. Thanks Roger for fixing this.
Subject: Bug 29132 Author: sayle Date: Thu Sep 21 17:13:33 2006 New Revision: 117114 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=117114 Log: 2006-09-21 Roger Sayle <roger@eyesopen.com> PR debug/29132 * dwarf2out.c (dwarf2out_begin_prologue): Initialise the current label, dw_fde_current_label, to be the start of the function, i.e. the same value as dw_fde_begin. Modified: trunk/gcc/ChangeLog trunk/gcc/dwarf2out.c
Fixed on mainline (confirmed on mips-sgi-irix6.5). It'll take another day or two to backport to the 4.1 branch, as bootstrap and regtest on MIPS takes a while.
Subject: Bug 29132 Author: sayle Date: Sun Sep 24 15:55:51 2006 New Revision: 117180 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=117180 Log: PR debug/29132 Backport from mainline * dwarf2out.c (dwarf2out_begin_prologue): Initialise the current label, dw_fde_current_label, to be the start of the function, i.e. the same value as dw_fde_begin. Modified: branches/gcc-4_1-branch/gcc/ChangeLog branches/gcc-4_1-branch/gcc/dwarf2out.c
Fixed.