This is the mail archive of the mailing list for the GDB project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: enable seperate debug info testing with stabs

Pedro Alves wrote:
The problems described in the comment I'm removing seem to
be solved already.  I can run the tests successfully on
Cygwin, where stabs is still the default.

Any comments on the patch below?

The stabs skipping came originally [1] from gdb.base/sepdebug.exp,
to address stabs+ problems reported here [2] and Daniel moved it later
to lib/gdb.exp [3].


To give it broader testing, I've tested it on i686-pc-linux-gnu/stabs+
and on i686-pc-cygwin/stabs+ with this board file, based on sepdebug.exp
and on a board file Jim posted on the list to use with elfutils.
This one's to use with binutils.  Its purpose is to run the
whole testsuite in "separate debug info" mode.

load_base_board_description "unix"
#load_base_board_description "cygwin"

proc ${current_target_name}_compile {source dest type options} {
    global strip_to_file_program
    global EXEEXT

    # Run the standard compilation procedure.
    set result [default_target_compile $source $dest $type $options]

    # If it didn't succeed, return directly.
    if {[string compare $result ""] != 0} {
        return $result

    if {[string compare $type executable] == 0} {
	# Note: the procedure gdb_gnu_strip_debug will produce an
	# executable called ${dest}, which is just like the
	# executable ($dest) but without the debuginfo.  Instead
	# $dest has a .gnudebuglink section which contains the name
	# of a debuginfo only file.  This file will be stored in the
	# .debug subdirectory.

	# Since strip and objdump don't add $EXEEXT automatically like
	# gcc does, gdb_gnu_strip_debug will fail when $dest doesn't
	# have $EXEEXT.  We workaround that by adding $EXEEXT here if
	# needed.  This is to avoid adding it throughout the
	# testsuite.
	set loc_dest $dest
	set exeext_len [string length $EXEEXT]
	set dest_len [string length $dest]
	if { $exeext_len > 0 && $dest_len > $exeext_len } {
	    set suf_pos [expr $dest_len - $exeext_len]
	    set suf [string range $dest $suf_pos $dest_len]
	    if {[string compare $EXEEXT $suf] != 0} {
		set loc_dest $dest$EXEEXT

	if [gdb_gnu_strip_debug ${loc_dest}] {
	    # check that you have a recent version of strip and
	    # objcopy installed
	    unsupported "cannot produce separate debug info files"
	    return -1

    return ""

proc ${board}_init { whole_name } {
    global board_info
    set baseboard [lindex [split $whole_name "/"] 0]
    set board_info($baseboard,isremote) 0


2007-10-29 Pedro Alves <>

* lib/gdb.exp (gdb_gnu_strip_debug): Remove debug format test.

 gdb/testsuite/lib/gdb.exp |   34 ----------------------------------
 1 file changed, 34 deletions(-)

Index: src/gdb/testsuite/lib/gdb.exp
--- src.orig/gdb/testsuite/lib/gdb.exp 2007-10-29 20:15:00.000000000 +0000
+++ src/gdb/testsuite/lib/gdb.exp 2007-10-29 20:23:32.000000000 +0000
@@ -2520,40 +2520,6 @@ proc build_id_debug_filename_get { exec proc gdb_gnu_strip_debug { dest args } {
- # First, make sure that we can do this. This is nasty. We need to
- # check for the stabs debug format. To do this we must run gdb on
- # the unstripped executable, list 'main' (as to have a default
- # source file), use get_debug_format (which does 'info source')
- # and then see if the debug info is stabs. If so, we bail out. We
- # cannot do this any other way because get_debug_format finds out
- # the debug format using gdb itself, and in case of stabs we get
- # an error loading the program if it is already stripped. An
- # alternative would be to find out the debug info from the flags
- # passed to dejagnu when the test is run.
- gdb_exit
- gdb_start
- gdb_load ${dest}
- gdb_test "list main" "" ""
- get_debug_format
- if { [test_debug_format "stabs"] } then {
- # The separate debug info feature doesn't work well in
- # binutils with stabs. It produces a corrupted debug info
- # only file, and gdb chokes on it. It is almost impossible to
- # capture the failing message out of gdb, because it happens
- # inside gdb_load. At that point any error message is
- # intercepted by dejagnu itself, and, because of the error
- # threshold, any faulty test result is changed into an
- # UNRESOLVED. (see dejagnu/lib/framework.exp)
- unsupported "no separate debug info handling with stabs"
- return -1
- } elseif { [test_debug_format "unknown"] } then {
- # gdb doesn't know what the debug format is. We are out of luck here. - unsupported "unknown debugging format"
- return -1
- }
- gdb_exit
set debug_file [separate_debug_filename $dest]
set strip_to_file_program [transform strip]
set objcopy_program [transform objcopy]

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