Letext and gdb problems on Solaris 8

Letext and gdb problems on Solaris 8

[Zartaj T. Majeed]

I am running into serious problems in gdb 5.0. Here are
the symptoms which occur at different places in the
program.

1. Step acts like next so I cannot step into a function
call.
2. gdb does not stop at breakpoints
3. bt shows only the current frame
4. bt shows frames down to an incorrectly labeled
function called Letext().

Any of these makes gdb very difficult to use but
encountering all of them at various points in the
program makes gdb virtually unusable.

I am running gdb 5.0 built with gcc 2.95.2 for
Solaris 8. My program is also built with gcc 2.95.2.

Trying gdb 5.1 did not help. When I ran the program
it got a SIGTRAP and I could not proceed.

The only clue I have is the presence of Letext in
certain stack traces. There is no such symbol in
my program and I know it is a symbol generated by gcc
but do not know its purpose--something like __etext?

Anyway bt in #4 shows something like:

#6 0x... in myfunc() at myfunc_file.c:100
#7 0x... in Letext() at funccall_file.c:80

And here's a disassembly of funccall_file.c:80 which
looks like some kind of stub. The actual function
body at that line is over 100 lines long, so this
is not its disassembly.

(gdb) disas
Dump of assembler code for function Letext:
0xfe06ffa8 <Letext>:    call  0xfe1a0794
0xfe06ffac <Letext+4>:  nop
0xfe06ffb0 <Letext+8>:  st  %o0, [ %fp + -20 ]
0xfe06ffb4 <Letext+12>: ld  [ %fp + -20 ], %o0
0xfe06ffb8 <Letext+16>: cmp  %o0, 0
0xfe06ffbc <Letext+20>: be  0xfe07015c <func1+52>
0xfe06ffc0 <Letext+24>: nop
0xfe06ffc4 <Letext+28>: clr  [ %fp + -40 ]
End of assembler dump.

If nobody has a solution, does anyone know how I can
go about determining what's going wrong with gdb
or code generated by gcc?

Thanks,
Zartaj