* RFA: test that GDB tolerates bad #inclusion data
@ 2003-09-23 16:13 Jim Blandy
[not found] ` <yf2he337a2e.fsf@hawaii.kealia.com>
0 siblings, 1 reply; 9+ messages in thread
From: Jim Blandy @ 2003-09-23 16:13 UTC (permalink / raw)
To: gdb-patches
2003-09-23 Jim Blandy <jimb@redhat.com>
* gdb.base/badinc.exp, gdb.base/badinc1.h, gdb.base/badinc2.h,
gdb.base/badinc3.h, gdb.base/badinc.c: New tests.
Index: gdb/testsuite/gdb.base/badinc.exp
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc.exp
diff -N gdb/testsuite/gdb.base/badinc.exp
*** gdb/testsuite/gdb.base/badinc.exp 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc.exp 23 Sep 2003 16:04:39 -0000
***************
*** 0 ****
--- 1,155 ----
+ # Test macro handling of #included files.
+ # Copyright 2003 Free Software Foundation, Inc.
+
+ # This program is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU General Public License as published by
+ # the Free Software Foundation; either version 2 of the License, or
+ # (at your option) any later version.
+ #
+ # This program is distributed in the hope that it will be useful,
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ # GNU General Public License for more details.
+ #
+ # You should have received a copy of the GNU General Public License
+ # along with this program; if not, write to the Free Software
+ # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+ # Please email any bugs, comments, and/or additions to this file to:
+ # bug-gdb@prep.ai.mit.edu
+
+ # Some versions of GCC produce Dwarf 2 macro information that lists
+ # all files #included via the '-imacros' flag as appearing at line 1
+ # of the main source file, like this:
+ #
+ # $ gcc -g3 -imacros badinc1.h -imacros badinc2.h badinc.c -o badinc
+ # $ readelf -wml badinc
+ # ...
+ # The File Name Table:
+ # Entry Dir Time Size Name
+ # 1 0 0 0 badinc.c
+ # 2 0 0 0 badinc1.h
+ # 3 0 0 0 badinc2.h
+ # 4 0 0 0 badinc3.h
+ # ...
+ # Contents of the .debug_macinfo section:
+ #
+ # DW_MACINFO_start_file - lineno: 0 filenum: 1
+ # DW_MACINFO_define - lineno : 1 macro : __VERSION__ "3.2 20020903 (Red Hat Linux 8.0 3.2-7)"
+ # DW_MACINFO_define - lineno : 2 macro : __USER_LABEL_PREFIX__
+ # ...
+ # DW_MACINFO_define - lineno : 1 macro : __i386__ 1
+ # DW_MACINFO_define - lineno : 1 macro : __tune_i386__ 1
+ # DW_MACINFO_start_file - lineno: 1 filenum: 2
+ # DW_MACINFO_define - lineno : 1 macro : FOO 1
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 2 filenum: 3
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 2
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 1 filenum: 4
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 3
+ # DW_MACINFO_end_file
+ # DW_MACINFO_end_file
+ # $
+ #
+ # Note how this attributes the inclusions of badinc1.h and badinc2.h
+ # (requested via the -imacros command-line switches) are attributed to
+ # lines 1 and 2 of badinc.c, and the #inclusion of badinc3.h
+ # (requested via a real #include directive) is attributed to line 1 of
+ # badinc.c.
+ #
+ # This script tests GDB's ability to work around this kind of bogus
+ # information. In particular, versions of GDB before Sep 2003 would
+ # get an internal error when trying to read the above debugging info.
+ #
+ # If more than one #inclusion is alleged to have taken place at the
+ # same source line, then GDB can't tell which file's definitions and
+ # undefinitions come first, so it can't tell which definitions are in
+ # scope after the #inclusions. To work around this, GDB puts all the
+ # files #included from a given source file in a list sorted by the
+ # line at which they were #included; this gives it the chance to
+ # detect multiple #inclusions at the same line, complain, and assign
+ # distinct, albiet incorrect, line numbers to each #inclusion.
+ #
+ # The bug was that GDB didn't put files in the sorted list correctly;
+ # the condition deciding how long to walk the list was reversed. This
+ # is why we have to have three #include files in the test: if there
+ # are only two headers, then when the second #inclusion at line 1
+ # (badinc3.h) is added to the list, the list has only one other
+ # element, and the traversal stops in the right place anyway. With
+ # three header files, badinc2.h, #included at line two, ends up before
+ # badinc1.h, #included at line one, so we don't notice the duplication
+ # when we insert badinc3.h at line one too.
+
+ if $tracelevel then {
+ strace $tracelevel
+ }
+
+ set prms_id 0
+ set bug_id 0
+
+ set testfile "badinc"
+ set binfile ${objdir}/${subdir}/${testfile}
+
+
+ # Create and source the file that provides information about the compiler
+ # used to compile the test case.
+ if [get_compiler_info ${binfile}] {
+ return -1;
+ }
+
+
+ if {$gcc_compiled} {
+ # Some versions of GCC (like 3.2) can be persuaded to produce
+ # appropriately bogus info by #including files using the -imacros
+ # flag.
+ set header1 "${srcdir}/${subdir}/badinc1.h"
+ set header2 "${srcdir}/${subdir}/badinc2.h"
+ set additional_flags \
+ "additional_flags=-imacros $header1 -imacros $header2"
+ if {[gdb_compile "${srcdir}/${subdir}/${testfile}.c" ${binfile} \
+ executable [list debug $additional_flags]] != ""} {
+ gdb_suppress_entire_file "Testcase compile failed, so all tests in this file will automatically fail."
+ }
+ } else {
+ # We don't know how to run this test on this compiler, so don't bother.
+ return 0
+ }
+
+ gdb_exit
+ gdb_start
+ gdb_reinitialize_dir $srcdir/$subdir
+ gdb_load ${binfile}
+
+ # Any command that causes GDB to read the debugging info for the
+ # badinc.c compilation unit will do here.
+ set test_name "tolerate macro info with bad #inclusion line numbers"
+ send_gdb "break main\n"
+ gdb_expect {
+ -re "Breakpoint 1 at 0x.*: file .*badinc.c.*\\.\r\n${gdb_prompt}" {
+ pass $test_name
+ }
+ -re ".*internal-error:.*.y or n. " {
+ fail $test_name
+ send_gdb "y\n"
+ gdb_expect {
+ -re ".*.y or n. " {
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "$gdb_prompt" {
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
+ }
+ -re ".*$gdb_prompt" {
+ fail "$test_name (unexpected response)"
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
Index: gdb/testsuite/gdb.base/badinc.c
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc.c
diff -N gdb/testsuite/gdb.base/badinc.c
*** gdb/testsuite/gdb.base/badinc.c 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc.c 23 Sep 2003 16:04:39 -0000
***************
*** 0 ****
--- 1,6 ----
+ #include "badinc3.h"
+
+ int
+ main (int argc, char **argv)
+ {
+ }
Index: gdb/testsuite/gdb.base/badinc1.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc1.h
diff -N gdb/testsuite/gdb.base/badinc1.h
*** gdb/testsuite/gdb.base/badinc1.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc1.h 23 Sep 2003 16:04:39 -0000
***************
*** 0 ****
--- 1 ----
+ #define FOO 1
Index: gdb/testsuite/gdb.base/badinc2.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc2.h
diff -N gdb/testsuite/gdb.base/badinc2.h
*** gdb/testsuite/gdb.base/badinc2.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc2.h 23 Sep 2003 16:04:39 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 2
Index: gdb/testsuite/gdb.base/badinc3.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc3.h
diff -N gdb/testsuite/gdb.base/badinc3.h
*** gdb/testsuite/gdb.base/badinc3.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc3.h 23 Sep 2003 16:04:39 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 3
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
[not found] ` <yf2he337a2e.fsf@hawaii.kealia.com>
@ 2003-09-23 17:06 ` Jim Blandy
0 siblings, 0 replies; 9+ messages in thread
From: Jim Blandy @ 2003-09-23 17:06 UTC (permalink / raw)
To: David Carlton; +Cc: gdb-patches
David Carlton <carlton@kealia.com> writes:
> On 23 Sep 2003 11:10:09 -0500, Jim Blandy <jimb@redhat.com> said:
>
> > + # Note how this attributes the inclusions of badinc1.h and badinc2.h
> > + # (requested via the -imacros command-line switches) are attributed to
> > + # lines 1 and 2 of badinc.c,
>
> I think the "this attributes" at the beginning should be deleted.
*moip*
2003-09-23 Jim Blandy <jimb@redhat.com>
* gdb.base/badinc.exp, gdb.base/badinc1.h, gdb.base/badinc2.h,
gdb.base/badinc3.h, gdb.base/badinc.c: New tests.
Index: gdb/testsuite/gdb.base/badinc.exp
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc.exp
diff -N gdb/testsuite/gdb.base/badinc.exp
*** gdb/testsuite/gdb.base/badinc.exp 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc.exp 23 Sep 2003 17:04:42 -0000
***************
*** 0 ****
--- 1,154 ----
+ # Test macro handling of #included files.
+ # Copyright 2003 Free Software Foundation, Inc.
+
+ # This program is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU General Public License as published by
+ # the Free Software Foundation; either version 2 of the License, or
+ # (at your option) any later version.
+ #
+ # This program is distributed in the hope that it will be useful,
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ # GNU General Public License for more details.
+ #
+ # You should have received a copy of the GNU General Public License
+ # along with this program; if not, write to the Free Software
+ # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+ # Please email any bugs, comments, and/or additions to this file to:
+ # bug-gdb@prep.ai.mit.edu
+
+ # Some versions of GCC produce Dwarf 2 macro information that lists
+ # all files #included via the '-imacros' flag as appearing at line 1
+ # of the main source file, like this:
+ #
+ # $ gcc -g3 -imacros badinc1.h -imacros badinc2.h badinc.c -o badinc
+ # $ readelf -wml badinc
+ # ...
+ # The File Name Table:
+ # Entry Dir Time Size Name
+ # 1 0 0 0 badinc.c
+ # 2 0 0 0 badinc1.h
+ # 3 0 0 0 badinc2.h
+ # 4 0 0 0 badinc3.h
+ # ...
+ # Contents of the .debug_macinfo section:
+ #
+ # DW_MACINFO_start_file - lineno: 0 filenum: 1
+ # DW_MACINFO_define - lineno : 1 macro : __VERSION__ "3.2 20020903 (Red Hat Linux 8.0 3.2-7)"
+ # DW_MACINFO_define - lineno : 2 macro : __USER_LABEL_PREFIX__
+ # ...
+ # DW_MACINFO_define - lineno : 1 macro : __i386__ 1
+ # DW_MACINFO_define - lineno : 1 macro : __tune_i386__ 1
+ # DW_MACINFO_start_file - lineno: 1 filenum: 2
+ # DW_MACINFO_define - lineno : 1 macro : FOO 1
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 2 filenum: 3
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 2
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 1 filenum: 4
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 3
+ # DW_MACINFO_end_file
+ # DW_MACINFO_end_file
+ # $
+ #
+ # Note how the inclusions of badinc1.h and badinc2.h (requested via
+ # the -imacros command-line switches) are attributed to lines 1 and 2
+ # of badinc.c, and the #inclusion of badinc3.h (requested via a real
+ # #include directive) is attributed to line 1 of badinc.c.
+ #
+ # This script tests GDB's ability to work around this kind of bogus
+ # information. In particular, versions of GDB before Sep 2003 would
+ # get an internal error when trying to read the above debugging info.
+ #
+ # If more than one #inclusion is alleged to have taken place at the
+ # same source line, then GDB can't tell which file's definitions and
+ # undefinitions come first, so it can't tell which definitions are in
+ # scope after the #inclusions. To work around this, GDB puts all the
+ # files #included from a given source file in a list sorted by the
+ # line at which they were #included; this gives it the chance to
+ # detect multiple #inclusions at the same line, complain, and assign
+ # distinct, albiet incorrect, line numbers to each #inclusion.
+ #
+ # The bug was that GDB didn't put files in the sorted list correctly;
+ # the condition deciding how long to walk the list was reversed. This
+ # is why we have to have three #include files in the test: if there
+ # are only two headers, then when the second #inclusion at line 1
+ # (badinc3.h) is added to the list, the list has only one other
+ # element, and the traversal stops in the right place anyway. With
+ # three header files, badinc2.h, #included at line two, ends up before
+ # badinc1.h, #included at line one, so we don't notice the duplication
+ # when we insert badinc3.h at line one too.
+
+ if $tracelevel then {
+ strace $tracelevel
+ }
+
+ set prms_id 0
+ set bug_id 0
+
+ set testfile "badinc"
+ set binfile ${objdir}/${subdir}/${testfile}
+
+
+ # Create and source the file that provides information about the compiler
+ # used to compile the test case.
+ if [get_compiler_info ${binfile}] {
+ return -1;
+ }
+
+
+ if {$gcc_compiled} {
+ # Some versions of GCC (like 3.2) can be persuaded to produce
+ # appropriately bogus info by #including files using the -imacros
+ # flag.
+ set header1 "${srcdir}/${subdir}/badinc1.h"
+ set header2 "${srcdir}/${subdir}/badinc2.h"
+ set additional_flags \
+ "additional_flags=-imacros $header1 -imacros $header2"
+ if {[gdb_compile "${srcdir}/${subdir}/${testfile}.c" ${binfile} \
+ executable [list debug $additional_flags]] != ""} {
+ gdb_suppress_entire_file "Testcase compile failed, so all tests in this file will automatically fail."
+ }
+ } else {
+ # We don't know how to run this test on this compiler, so don't bother.
+ return 0
+ }
+
+ gdb_exit
+ gdb_start
+ gdb_reinitialize_dir $srcdir/$subdir
+ gdb_load ${binfile}
+
+ # Any command that causes GDB to read the debugging info for the
+ # badinc.c compilation unit will do here.
+ set test_name "tolerate macro info with bad #inclusion line numbers"
+ send_gdb "break main\n"
+ gdb_expect {
+ -re "Breakpoint 1 at 0x.*: file .*badinc.c.*\\.\r\n${gdb_prompt}" {
+ pass $test_name
+ }
+ -re ".*internal-error:.*.y or n. " {
+ fail $test_name
+ send_gdb "y\n"
+ gdb_expect {
+ -re ".*.y or n. " {
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "$gdb_prompt" {
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
+ }
+ -re ".*$gdb_prompt" {
+ fail "$test_name (unexpected response)"
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
Index: gdb/testsuite/gdb.base/badinc.c
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc.c
diff -N gdb/testsuite/gdb.base/badinc.c
*** gdb/testsuite/gdb.base/badinc.c 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc.c 23 Sep 2003 17:04:42 -0000
***************
*** 0 ****
--- 1,6 ----
+ #include "badinc3.h"
+
+ int
+ main (int argc, char **argv)
+ {
+ }
Index: gdb/testsuite/gdb.base/badinc1.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc1.h
diff -N gdb/testsuite/gdb.base/badinc1.h
*** gdb/testsuite/gdb.base/badinc1.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc1.h 23 Sep 2003 17:04:42 -0000
***************
*** 0 ****
--- 1 ----
+ #define FOO 1
Index: gdb/testsuite/gdb.base/badinc2.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc2.h
diff -N gdb/testsuite/gdb.base/badinc2.h
*** gdb/testsuite/gdb.base/badinc2.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc2.h 23 Sep 2003 17:04:42 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 2
Index: gdb/testsuite/gdb.base/badinc3.h
===================================================================
RCS file: gdb/testsuite/gdb.base/badinc3.h
diff -N gdb/testsuite/gdb.base/badinc3.h
*** gdb/testsuite/gdb.base/badinc3.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/badinc3.h 23 Sep 2003 17:04:42 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 3
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
2003-09-24 18:13 ` Jim Blandy
2003-09-24 18:13 ` Jim Blandy
@ 2003-09-24 22:26 ` Jim Blandy
1 sibling, 0 replies; 9+ messages in thread
From: Jim Blandy @ 2003-09-24 22:26 UTC (permalink / raw)
To: Michael Elizabeth Chastain; +Cc: carlton, gdb-patches
Jim Blandy <jimb@redhat.com> writes:
> I'm glad you asked me to revise this:
>
> - The test is no lenger GCC 3.2 specific; GCC 3.3 will also produce
> macro information that will crash GDB without the corresponding
> patch to macrotab.c.
>
> - The test is no longer GCC-specific at all. I have no idea what
> other compilers will generate for that test program, but something
> reasonable will happen.
>
> It now uses gdb_test_multiple for the outermost test. I tried using
> it for the internal error recovery too, but it got kind of upset and
> produced extraneous ERRORS and UNRESOLVED tests.
Actually, that was an intermediate revision; here's the final version:
2003-09-23 Jim Blandy <jimb@redhat.com>
* gdb.base/lineinc.exp, gdb.base/lineinc1.h, gdb.base/lineinc2.h,
gdb.base/lineinc3.h, gdb.base/lineinc.c: New tests.
Index: gdb/testsuite/gdb.base/lineinc.exp
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc.exp
diff -N gdb/testsuite/gdb.base/lineinc.exp
*** gdb/testsuite/gdb.base/lineinc.exp 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc.exp 24 Sep 2003 17:35:19 -0000
***************
*** 0 ****
--- 1,126 ----
+ # Test macro handling of #included files.
+ # Copyright 2003 Free Software Foundation, Inc.
+
+ # This program is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU General Public License as published by
+ # the Free Software Foundation; either version 2 of the License, or
+ # (at your option) any later version.
+ #
+ # This program is distributed in the hope that it will be useful,
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ # GNU General Public License for more details.
+ #
+ # You should have received a copy of the GNU General Public License
+ # along with this program; if not, write to the Free Software
+ # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+ # Please email any bugs, comments, and/or additions to this file to:
+ # bug-gdb@prep.ai.mit.edu
+
+ # The test program lineinc.c contains a mix of #line directives and
+ # #include directives that will cause the compiler to attribute more
+ # than one #inclusion to the same source line. You can get similar
+ # effects using things like GCC's '-imacros' flag.
+ #
+ # Compiling lineinc.c with Dwarf 2 macro information will produce
+ # something like this:
+ #
+ # $ gcc -g3 lineinc.c -o lineinc
+ # $ readelf -wml lineinc
+ # ...
+ # The File Name Table:
+ # Entry Dir Time Size Name
+ # 1 0 0 0 lineinc.c
+ # 2 0 0 0 lineinc1.h
+ # 3 0 0 0 lineinc2.h
+ # 4 0 0 0 lineinc3.h
+ # ...
+ # Contents of the .debug_macinfo section:
+ #
+ # DW_MACINFO_start_file - lineno: 0 filenum: 1
+ # DW_MACINFO_define - lineno : 1 macro : __VERSION__ "3.2 20020903 (Red Hat Linux 8.0 3.2-7)"
+ # DW_MACINFO_define - lineno : 2 macro : __USER_LABEL_PREFIX__
+ # ...
+ # DW_MACINFO_define - lineno : 1 macro : __i386__ 1
+ # DW_MACINFO_define - lineno : 1 macro : __tune_i386__ 1
+ # DW_MACINFO_start_file - lineno: 10 filenum: 2
+ # DW_MACINFO_define - lineno : 1 macro : FOO 1
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 10 filenum: 3
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 2
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 11 filenum: 4
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 3
+ # DW_MACINFO_end_file
+ # DW_MACINFO_end_file
+ # $
+ #
+ # Note how the inclusions of lineinc1.h and lineinc2.h are both
+ # attributed to line 10 of lineinc.c, and the #inclusion of lineinc3.h
+ # is attributed to line 11. This is all correct, given the #line
+ # directives in lineinc.c.
+ #
+ # Dwarf 2 macro information doesn't contain enough information to
+ # allow GDB to figure out what's really going on here --- it makes no
+ # mention of the #line directives --- so we just try to cope as best
+ # we can. If the macro table were to attribute more than one
+ # #inclusion to the same source line, then GDB wouldn't be able to
+ # tell which #included file's #definitions and #undefinitions come
+ # first, so it can't tell which #definitions are in scope following
+ # all the #inclusions. To cope with this, GDB puts all the files
+ # #included by a given source file in a list sorted by the line at
+ # which they were #included; this gives GDB the chance to detect
+ # multiple #inclusions at the same line, complain, and assign
+ # distinct, albiet incorrect, line numbers to each #inclusion.
+ #
+ # However, at one point GDB was sorting the list in reverse order,
+ # while the code to assign new, distinct line numbers assumed it was
+ # sorted in ascending order; GDB would get an internal error trying to
+ # read the above debugging info.
+
+ if $tracelevel then {
+ strace $tracelevel
+ }
+
+ set prms_id 0
+ set bug_id 0
+
+ set testfile "lineinc"
+ set binfile ${objdir}/${subdir}/${testfile}
+
+
+ if {[gdb_compile "${srcdir}/${subdir}/${testfile}.c" ${binfile} executable {debug}] != ""} {
+ gdb_suppress_entire_file "Testcase compile failed, so all tests in this file will automatically fail."
+ }
+
+ gdb_exit
+ gdb_start
+ gdb_reinitialize_dir $srcdir/$subdir
+ gdb_load ${binfile}
+
+ # Any command that causes GDB to read the debugging info for the
+ # lineinc.c compilation unit will do here.
+ set test_name "tolerate macro info with multiple #inclusions per line"
+ gdb_test_multiple "break main" $test_name {
+ -re "Breakpoint 1 at 0x.*: file .*lineinc.c.*\\.\r\n${gdb_prompt}" {
+ pass $test_name
+ }
+ -re ".*internal-error:.*.y or n. " {
+ fail $test_name
+ send_gdb "y\n"
+ gdb_expect {
+ -re ".*.y or n. " {
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "$gdb_prompt" {
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
+ }
+ }
Index: gdb/testsuite/gdb.base/lineinc.c
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc.c
diff -N gdb/testsuite/gdb.base/lineinc.c
*** gdb/testsuite/gdb.base/lineinc.c 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc.c 24 Sep 2003 17:35:19 -0000
***************
*** 0 ****
--- 1,30 ----
+ /* The following is written to tickle a specific bug in the macro
+ table code (now hopefully fixed), which doesn't insert new included
+ files in the #including file's list in the proper place. They
+ should be sorted by the number of the line which #included them, in
+ increasing order, but the sense of the comparison was reversed, so
+ the list ends up being built backwards. This isn't a problem by
+ itself, but the code to pick new, non-conflicting line numbers for
+ headers alleged to be #included at the same line as some other
+ header assumes that the list's line numbers are in ascending order.
+
+ So, given the following input, lineinc1.h gets added to lineinc.c's
+ #inclusion list first, at line 10. When the debug info reader
+ tries to add lineinc2.h at line 10 as well, the code will notice the
+ duplication --- since there's only one extant element in the list,
+ it'll find it --- and insert it after lineinc1.h, with line 11.
+ Since the code is putting the list in order of descending
+ #inclusion line number, the list is now out of order. When we try
+ to #include lineinc3.h at line 11, we won't notice the duplication. */
+
+ #line 10
+ #include "lineinc1.h"
+ #line 10
+ #include "lineinc2.h"
+ #line 11
+ #include "lineinc3.h"
+
+ int
+ main (int argc, char **argv)
+ {
+ }
Index: gdb/testsuite/gdb.base/lineinc1.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc1.h
diff -N gdb/testsuite/gdb.base/lineinc1.h
*** gdb/testsuite/gdb.base/lineinc1.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc1.h 24 Sep 2003 17:35:19 -0000
***************
*** 0 ****
--- 1 ----
+ #define FOO 1
Index: gdb/testsuite/gdb.base/lineinc2.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc2.h
diff -N gdb/testsuite/gdb.base/lineinc2.h
*** gdb/testsuite/gdb.base/lineinc2.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc2.h 24 Sep 2003 17:35:19 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 2
Index: gdb/testsuite/gdb.base/lineinc3.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc3.h
diff -N gdb/testsuite/gdb.base/lineinc3.h
*** gdb/testsuite/gdb.base/lineinc3.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc3.h 24 Sep 2003 17:35:19 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 3
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
2003-09-23 22:33 ` Jim Blandy
@ 2003-09-24 18:13 ` Jim Blandy
2003-09-24 18:13 ` Jim Blandy
2003-09-24 22:26 ` Jim Blandy
0 siblings, 2 replies; 9+ messages in thread
From: Jim Blandy @ 2003-09-24 18:13 UTC (permalink / raw)
To: Michael Elizabeth Chastain; +Cc: carlton, gdb-patches
I'm glad you asked me to revise this:
- The test is no lenger GCC 3.2 specific; GCC 3.3 will also produce
macro information that will crash GDB without the corresponding
patch to macrotab.c.
- The test is no longer GCC-specific at all. I have no idea what
other compilers will generate for that test program, but something
reasonable will happen.
It now uses gdb_test_multiple for the outermost test. I tried using
it for the internal error recovery too, but it got kind of upset and
produced extraneous ERRORS and UNRESOLVED tests.
2003-09-23 Jim Blandy <jimb@redhat.com>
* gdb.base/lineinc.exp, gdb.base/lineinc1.h, gdb.base/lineinc2.h,
gdb.base/lineinc3.h, gdb.base/lineinc.c: New tests.
Index: gdb/testsuite/gdb.base/lineinc.exp
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc.exp
diff -N gdb/testsuite/gdb.base/lineinc.exp
*** gdb/testsuite/gdb.base/lineinc.exp 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc.exp 24 Sep 2003 17:13:19 -0000
***************
*** 0 ****
--- 1,133 ----
+ # Test macro handling of #included files.
+ # Copyright 2003 Free Software Foundation, Inc.
+
+ # This program is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU General Public License as published by
+ # the Free Software Foundation; either version 2 of the License, or
+ # (at your option) any later version.
+ #
+ # This program is distributed in the hope that it will be useful,
+ # but WITHOUT ANY WARRANTY; without even the implied warranty of
+ # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ # GNU General Public License for more details.
+ #
+ # You should have received a copy of the GNU General Public License
+ # along with this program; if not, write to the Free Software
+ # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+
+ # Please email any bugs, comments, and/or additions to this file to:
+ # bug-gdb@prep.ai.mit.edu
+
+ # The test program lineinc.c contains a mix of #line directives and
+ # #include directives that will cause the compiler to attribute more
+ # than one #inclusion to the same source line. You can get similar
+ # effects using things like GCC's '-imacros' flag.
+ #
+ # Compiling lineinc.c with Dwarf 2 macro information will produce
+ # something like this:
+ #
+ # $ gcc -g3 lineinc.c -o lineinc
+ # $ readelf -wml lineinc
+ # ...
+ # The File Name Table:
+ # Entry Dir Time Size Name
+ # 1 0 0 0 lineinc.c
+ # 2 0 0 0 lineinc1.h
+ # 3 0 0 0 lineinc2.h
+ # 4 0 0 0 lineinc3.h
+ # ...
+ # Contents of the .debug_macinfo section:
+ #
+ # DW_MACINFO_start_file - lineno: 0 filenum: 1
+ # DW_MACINFO_define - lineno : 1 macro : __VERSION__ "3.2 20020903 (Red Hat Linux 8.0 3.2-7)"
+ # DW_MACINFO_define - lineno : 2 macro : __USER_LABEL_PREFIX__
+ # ...
+ # DW_MACINFO_define - lineno : 1 macro : __i386__ 1
+ # DW_MACINFO_define - lineno : 1 macro : __tune_i386__ 1
+ # DW_MACINFO_start_file - lineno: 10 filenum: 2
+ # DW_MACINFO_define - lineno : 1 macro : FOO 1
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 10 filenum: 3
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 2
+ # DW_MACINFO_end_file
+ # DW_MACINFO_start_file - lineno: 11 filenum: 4
+ # DW_MACINFO_undef - lineno : 1 macro : FOO
+ # DW_MACINFO_define - lineno : 2 macro : FOO 3
+ # DW_MACINFO_end_file
+ # DW_MACINFO_end_file
+ # $
+ #
+ # Note how the inclusions of lineinc1.h and lineinc2.h are both
+ # attributed to line 10 of lineinc.c, and the #inclusion of lineinc3.h
+ # is attributed to line 11. This is all correct, given the #line
+ # directives in lineinc.c.
+ #
+ # Dwarf 2 macro information doesn't contain enough information to
+ # allow GDB to figure out what's really going on here --- it makes no
+ # mention of the #line directives --- so we just try to cope as best
+ # we can. If the macro table were to attribute more than one
+ # #inclusion to the same source line, then GDB wouldn't be able to
+ # tell which #included file's #definitions and #undefinitions come
+ # first, so it can't tell which #definitions are in scope following
+ # all the #inclusions. To cope with this, GDB puts all the files
+ # #included by a given source file in a list sorted by the line at
+ # which they were #included; this gives GDB the chance to detect
+ # multiple #inclusions at the same line, complain, and assign
+ # distinct, albiet incorrect, line numbers to each #inclusion.
+ #
+ # However, at one point GDB was sorting the list in reverse order,
+ # while the code to assign new, distinct line numbers assumed it was
+ # sorted in ascending order; GDB would get an internal error trying to
+ # read the above debugging info.
+
+ if $tracelevel then {
+ strace $tracelevel
+ }
+
+ set prms_id 0
+ set bug_id 0
+
+ set testfile "lineinc"
+ set binfile ${objdir}/${subdir}/${testfile}
+
+
+ if {[gdb_compile "${srcdir}/${subdir}/${testfile}.c" ${binfile} executable {debug}] != ""} {
+ gdb_suppress_entire_file "Testcase compile failed, so all tests in this file will automatically fail."
+ }
+
+ gdb_exit
+ gdb_start
+ gdb_reinitialize_dir $srcdir/$subdir
+ gdb_load ${binfile}
+
+ # Any command that causes GDB to read the debugging info for the
+ # lineinc.c compilation unit will do here.
+ set test_name "tolerate macro info with multiple #inclusions per line"
+ send_gdb "break main\n"
+ gdb_expect {
+ -re "Breakpoint 1 at 0x.*: file .*lineinc.c.*\\.\r\n${gdb_prompt}" {
+ pass $test_name
+ }
+ -re ".*internal-error:.*.y or n. " {
+ fail $test_name
+ send_gdb "y\n"
+ gdb_expect {
+ -re ".*.y or n. " {
+ send_gdb "n\n"
+ exp_continue
+ }
+ -re "$gdb_prompt" {
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
+ }
+ -re ".*$gdb_prompt" {
+ fail "$test_name (unexpected response)"
+ }
+ timeout {
+ fail "$test_name (timeout)"
+ }
+ }
Index: gdb/testsuite/gdb.base/lineinc.c
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc.c
diff -N gdb/testsuite/gdb.base/lineinc.c
*** gdb/testsuite/gdb.base/lineinc.c 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc.c 24 Sep 2003 17:13:19 -0000
***************
*** 0 ****
--- 1,30 ----
+ /* The following is written to tickle a specific bug in the macro
+ table code (now hopefully fixed), which doesn't insert new included
+ files in the #including file's list in the proper place. They
+ should be sorted by the number of the line which #included them, in
+ increasing order, but the sense of the comparison was reversed, so
+ the list ends up being built backwards. This isn't a problem by
+ itself, but the code to pick new, non-conflicting line numbers for
+ headers alleged to be #included at the same line as some other
+ header assumes that the list's line numbers are in ascending order.
+
+ So, given the following input, lineinc1.h gets added to lineinc.c's
+ #inclusion list first, at line 10. When the debug info reader
+ tries to add lineinc2.h at line 10 as well, the code will notice the
+ duplication --- since there's only one extant element in the list,
+ it'll find it --- and insert it after lineinc1.h, with line 11.
+ Since the code is putting the list in order of descending
+ #inclusion line number, the list is now out of order. When we try
+ to #include lineinc3.h at line 11, we won't notice the duplication. */
+
+ #line 10
+ #include "lineinc1.h"
+ #line 10
+ #include "lineinc2.h"
+ #line 11
+ #include "lineinc3.h"
+
+ int
+ main (int argc, char **argv)
+ {
+ }
Index: gdb/testsuite/gdb.base/lineinc1.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc1.h
diff -N gdb/testsuite/gdb.base/lineinc1.h
*** gdb/testsuite/gdb.base/lineinc1.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc1.h 24 Sep 2003 17:13:19 -0000
***************
*** 0 ****
--- 1 ----
+ #define FOO 1
Index: gdb/testsuite/gdb.base/lineinc2.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc2.h
diff -N gdb/testsuite/gdb.base/lineinc2.h
*** gdb/testsuite/gdb.base/lineinc2.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc2.h 24 Sep 2003 17:13:19 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 2
Index: gdb/testsuite/gdb.base/lineinc3.h
===================================================================
RCS file: gdb/testsuite/gdb.base/lineinc3.h
diff -N gdb/testsuite/gdb.base/lineinc3.h
*** gdb/testsuite/gdb.base/lineinc3.h 1 Jan 1970 00:00:00 -0000
--- gdb/testsuite/gdb.base/lineinc3.h 24 Sep 2003 17:13:19 -0000
***************
*** 0 ****
--- 1,2 ----
+ #undef FOO
+ #define FOO 3
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
2003-09-24 18:13 ` Jim Blandy
@ 2003-09-24 18:13 ` Jim Blandy
2003-09-24 22:26 ` Jim Blandy
1 sibling, 0 replies; 9+ messages in thread
From: Jim Blandy @ 2003-09-24 18:13 UTC (permalink / raw)
To: Michael Elizabeth Chastain; +Cc: carlton, gdb-patches
Jim Blandy <jimb@redhat.com> writes:
> I'm glad you asked me to revise this:
Oh, and re-doing it helped me realize that it's not a case of GDB
coping with bogus information from the compiler --- a correct test
program, with plausible macro debug info, could also trigger the bug.
Thus the test being renamed from 'badinc' to 'lineinc'.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
@ 2003-09-24 3:42 Michael Elizabeth Chastain
0 siblings, 0 replies; 9+ messages in thread
From: Michael Elizabeth Chastain @ 2003-09-24 3:42 UTC (permalink / raw)
To: jimb; +Cc: carlton, gdb-patches
jimb> Um, why would I want to bow to a charismatic leader's will and commit
jimb> an act of ultimate self-destruction?
Didn't you get a packet of kool-aid in the mail along with
the copyright assignment forms? :)
Michael C
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
2003-09-23 18:03 Michael Elizabeth Chastain
@ 2003-09-23 22:33 ` Jim Blandy
2003-09-24 18:13 ` Jim Blandy
0 siblings, 1 reply; 9+ messages in thread
From: Jim Blandy @ 2003-09-23 22:33 UTC (permalink / raw)
To: Michael Elizabeth Chastain; +Cc: carlton, gdb-patches
Michael Elizabeth Chastain <mec@shout.net> writes:
> Jim, we gotta feed you the gdb_test_multiple kool-aid.
> See gdb1250.exp or float.exp or watchpoint.exp.
Um, why would I want to bow to a charismatic leader's will and commit
an act of ultimate self-destruction?
> Also, could you avoid the dependency on gcc version by using
> preprocessed ".i" files with the bad line numbers already in them?
> Instead of / in addition to the tests with C source.
Let me fool around a bit with #line directives and see if I can do as
well without the -imacro switches.
In the end, though, it's up to the compiler to generate correct
Dwarf 2 information, so it'll always be compiler-dependent.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
@ 2003-09-23 19:59 Michael Elizabeth Chastain
0 siblings, 0 replies; 9+ messages in thread
From: Michael Elizabeth Chastain @ 2003-09-23 19:59 UTC (permalink / raw)
To: carlton, jimb; +Cc: gdb-patches
It works in my test bed. All tests PASSed in all configurations.
gdb HEAD
gcc 2.95.3, gcc 3.2-7-rh, gcc 3.3.1, gcc HEAD
binutils 2.13.90.2-2, binutils 2.14
dwarf-2, stabs+
-g2, -g3
gdb HEAD is 2003-09-21 04:14:29 UTC (approximately)
gcc HEAD is 2003-09-23 17:57:08 UTC (approximately)
I did have other objections: gdb_test_multiple, and writing the
test as a ".i" file to avoid gcc dependencies.
Michael C
===
2003-09-23 Jim Blandy <jimb@redhat.com>
* gdb.base/badinc.exp, gdb.base/badinc1.h, gdb.base/badinc2.h,
gdb.base/badinc3.h, gdb.base/badinc.c: New tests.
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: RFA: test that GDB tolerates bad #inclusion data
@ 2003-09-23 18:03 Michael Elizabeth Chastain
2003-09-23 22:33 ` Jim Blandy
0 siblings, 1 reply; 9+ messages in thread
From: Michael Elizabeth Chastain @ 2003-09-23 18:03 UTC (permalink / raw)
To: carlton, jimb; +Cc: gdb-patches
Jim, we gotta feed you the gdb_test_multiple kool-aid.
See gdb1250.exp or float.exp or watchpoint.exp.
Also, could you avoid the dependency on gcc version by using
preprocessed ".i" files with the bad line numbers already in them?
Instead of / in addition to the tests with C source.
I will run this patch through my testbed with -g2 and -g3
(I do have -g3 capability, I just don't do it in the weeklies yet)
and will post results in an hour or two.
Michael C
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2003-09-24 22:26 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-09-23 16:13 RFA: test that GDB tolerates bad #inclusion data Jim Blandy
[not found] ` <yf2he337a2e.fsf@hawaii.kealia.com>
2003-09-23 17:06 ` Jim Blandy
2003-09-23 18:03 Michael Elizabeth Chastain
2003-09-23 22:33 ` Jim Blandy
2003-09-24 18:13 ` Jim Blandy
2003-09-24 18:13 ` Jim Blandy
2003-09-24 22:26 ` Jim Blandy
2003-09-23 19:59 Michael Elizabeth Chastain
2003-09-24 3:42 Michael Elizabeth Chastain
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox