RFA: test that GDB tolerates bad #inclusion data

Jim Blandy jimb@redhat.com
Wed Sep 24 18:13:00 GMT 2003


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



More information about the Gdb-patches mailing list