This is the mail archive of the gas2@sourceware.cygnus.com mailing list for the gas2 project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

gas dumps core


Here is all I know with a test case.

Snapshot:      gas-960503
Configuration: m68k-coff

Test file (Yes it is the one line starting with a "#"):

# 1 ""

gdb trace:

Breakpoint 9 at 0xef76c1b8
Program received signal SIGSEGV, Segmentation fault.
0x29d84 in yank_symbols () at obj-format.c:2730
2730                  string_byte_count += strlen 
(filename_list_scan->filename) + 1;
(gdb) 

Further investigation shows that filename_list_scan is NULL. 

I don't really know the proper way to address this.  I don't know coff and
this is a pretty complicated routine. 

A simple fix would be to recognize that filename_list_scan is NULL and
avoid the strlen.  However, based on the comment "If the filename was too 
long to fit in the auxent, put it in the string table" it would appear 
that some code before here got tricked.

Does anyone more knowledgeable about coff have a suggestion?

Thanks.

--joel