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]

[PATCH] Fix regression in -file-list-exec-source-files command.

This is to fix one of the issues that was blocking for the 7.2 release :).

See for
a description of the problem. Namely, the file and fullname
fields are inverted in the output of the -file-list-exec-source-files
GDB/MI command:

    (gdb) interpreter-exec mi -file-list-exec-source-files

It turns out to be a silly thinko: The map_symbol_filenames function
calls the psymtab version of map_symbol_filenames routine, and this
version called the callback function with filename and fullname
in the wrong order (fullname/filename instead of filename/fullname).

The routine description in symfile.h confirst that expected order for
the FUN callback parameters:

    /* Call a callback for every file defined in OBJFILE.  FUN is the
       callback.  It is passed the file's name, the file's full name,
       and the DATA passed to this function.  */
    void (*map_symbol_filenames) (struct objfile *objfile,
                                  void (*fun) (const char *, const char *,
                                               void *),
                                  void *data);

Fixing this error uncovered another location where the arguments
were reversed: maybe_add_partial_symtab_filename.  Once the first
error was fixed, the debugger would crash while attempting to do
completion, because it was given a NULL fullname instead of the
non-NULL filename.

IMO, we would had a better chance of avoiding this sort of issue if we
had named the parameters in the callback function. It was less than
immediate to verify the order of arguments when I found the source of
the first problem (unnamed parameters, and lack of function documentation).
What I suggest is to create a type for the function pointer callback,
and have the parameters named and documented... But that can be taken
care of through another patch.


        * psymtab.c (map_symbol_filenames_psymtab): Call FUN with
        the arguments in the correct order.
        * symtab.c (maybe_add_partial_symtab_filename): Declare
        the arguments in the correct order.

I've checked this in HEAD and 7.2 branch, as it seems sufficiently obvious.
This was tested on x86_64-linux in HEAD.  I'll see whether it's possible
to create a regression test or not, next.

 gdb/ChangeLog |    7 +++++++
 gdb/psymtab.c |    2 +-
 gdb/symtab.c  |    2 +-
 3 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index b889876..580be2b 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,10 @@
+2010-08-19  Joel Brobecker  <>
+	* psymtab.c (map_symbol_filenames_psymtab): Call FUN with
+	the arguments in the correct order.
+	* symtab.c (maybe_add_partial_symtab_filename): Declare
+	the arguments in the correct order.
 2010-08-19  Jan Kratochvil  <>
 	* varobj.c (varobj_create): Replace variable old_fi with old_id,
diff --git a/gdb/psymtab.c b/gdb/psymtab.c
index bc47681..228b7a8 100644
--- a/gdb/psymtab.c
+++ b/gdb/psymtab.c
@@ -913,7 +913,7 @@ map_symbol_filenames_psymtab (struct objfile *objfile,
       fullname = psymtab_to_fullname (ps);
-      (*fun) (fullname, ps->filename, data);
+      (*fun) (ps->filename, fullname, data);
diff --git a/gdb/symtab.c b/gdb/symtab.c
index d43d573..baf6d94 100644
--- a/gdb/symtab.c
+++ b/gdb/symtab.c
@@ -4072,7 +4072,7 @@ struct add_partial_filename_data
 /* A callback for map_partial_symbol_filenames.  */
 static void
-maybe_add_partial_symtab_filename (const char *fullname, const char *filename,
+maybe_add_partial_symtab_filename (const char *filename, const char *fullname,
 				   void *user_data)
   struct add_partial_filename_data *data = user_data;

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