This is the mail archive of the gdb-patches@sources.redhat.com 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: -file-list-exec-source-files


Elena, thanks for taking the time to review my patch.

When is a good point to resubmit a patch? After every review, or after
all the issues are ironed out?

>  > Here is an initial patch for -file-list-exec-source-files.
>  > Some feedback would be appreciated.
>  > 
>  > I ran the testsuite and the results are the same before and after this
>  > patch.
>  > 
>  > Index: gdb/ChangeLog
>  > 	* dbxread.c (read_dbx_symtab): set pst->dirname when known
> 
> Each entry should start with capital letter and end with period.
> 
> I see some coding standards are not adhered to throughout the code.
> Most noticeably "foo ( int a )" should be "foo (int a)". Similarly for
> calls.

Ok, I will try to fix all of the coding standard errors. Is there a
program I can run on the source files before I create the diff that
formats the code according to the standard?

> 
>  > 
>  > Index: gdb/dbxread.c
>  > ===================================================================
>  > Index: gdb/dwarf2read.c
>  > ===================================================================
> 
> These are ok

Great!

>  >  
>  > Index: gdb/defs.h
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/defs.h,v
>  > retrieving revision 1.143
>  > diff -w -u -r1.143 defs.h
>  > --- gdb/defs.h	23 Feb 2004 19:26:14 -0000	1.143
>  > +++ gdb/defs.h	25 Feb 2004 03:51:35 -0000
>  > @@ -616,8 +616,6 @@
>  >  
>  >  extern void init_last_source_visited (void);
>  >  
>  > -extern char *symtab_to_filename (struct symtab *);
>  > -
>  >  /* From exec.c */
>  >  
>  >  extern void exec_set_section_offsets (bfd_signed_vma text_off,
>  > Index: gdb/dwarf2read.c
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/dwarf2read.c,v
>  > retrieving revision 1.135
>  > diff -w -u -r1.135 dwarf2read.c
>  > --- gdb/dwarf2read.c	21 Feb 2004 02:13:35 -0000	1.135
>  > +++ gdb/dwarf2read.c	25 Feb 2004 03:51:43 -0000
>  > @@ -316,6 +316,7 @@
>  >      unsigned int offset;
>  >      unsigned int abbrev;
>  >      char *name;
>  > +    char *dirname;
>  >      int has_pc_info;
>  >      CORE_ADDR lowpc;
>  >      CORE_ADDR highpc;
>  > @@ -1254,6 +1255,8 @@
>  >  				  objfile->global_psymbols.next,
>  >  				  objfile->static_psymbols.next);
>  >  
>  > +      pst->dirname = xstrdup ( comp_unit_die.dirname );
>  > +
>  >        pst->read_symtab_private = (char *)
>  >  	obstack_alloc (&objfile->objfile_obstack, sizeof (struct dwarf2_pinfo));
>  >        DWARF_INFO_BUFFER (pst) = dwarf_info_buffer;
>  > @@ -4326,6 +4329,10 @@
>  >  	  /* Prefer DW_AT_MIPS_linkage_name over DW_AT_name.  */
>  >  	  if (part_die->name == NULL)
>  >  	    part_die->name = DW_STRING (&attr);
>  > +	  break;
>  > +	case DW_AT_comp_dir:
>  > +	  if (part_die->dirname == NULL)
>  > +	    part_die->dirname = DW_STRING (&attr);
> 
> The dwarf2 specs say that the name is in the form ":pathname" or
> "hostname:pathname". Should we worry about the hostname? Does gcc emit
> that?  I have looked at a few executables and didn't see the hostname
> part.

It would probably just be best if you told me what case's you want me to
implement here. It seems that Jason Molenda understodd most of the
cases. I really don't know anything about what GCC emits and would would
be practical to implement.

>  > Index: gdb/source.c
>  > ===================================================================
> 
> this part I am not clear about.

Ok, Ok. I thought about this a lot. I think I made the best decision and
can describe why.

A few assumptions are in order. In order to get the fullname (abs path) 
to a file, GDB need's three things. The directory the file was compiled 
in (dirname), the filename in question (filename) and a list of paths 
to search.

> There is already a function called source_full_path_of() would it help
> if you used it?

The function source_full_path_of does not take into account 'dirname'.
It calls openp, which is not capable of finding the fullname of a file,
since it doesn't understand what dirname is. Basically, I don't even
think this function (source_full_path_of) is "truly" capable of 
finding the fullpath to a file. However, instead of removing it, 
I left it, since caller's of this function might be using for something 
I know nothing about.

> What is the difference between find_and_open_source and
> open_source_file?  I.e. why did you need to introduce it. I think it's
> not clear just from your comments about the file possibly baing moved
> around.

open_source_file was left around for backwards compatibility. The unit
source.c was used to calling a function, with just passing the symtab,
and getting back the symtab with a valid fullname. I could remove all
occurences of this function and replace it with symtab_to_fullname.

> I am a bit worried about the substitution of symtab_to_filename with
> symtab_to_fullname. The former returns null only if there is no
> symtab.  The latter returns null when there is no symtab OR when it
> cannot find the file. So the behavior is slightly different.

I basically think that the call -file-list-exec-source-files shouldn't 
'cache' it's results. GDB looks for each file, every time it is
requested to get the fullname. This is because, the user could have 
changed the path, or moved/deleted the file. I don't think GDB should
just return the filename instead, of the fullname.

So, if find_and_open_source couldn't "find and open the source file", it
returns NULL. Also, as a side effect, fullname in the [ps]ymtab also
get's set to NULL.

The testsuite didn't seem to have a problem with this, and I think it
makes sense to not trick the caller into having results when it couldn't
find any.

If the caller really wanted this functionality,
  return s->filename; /* File not found.  Just use short name */
I believe it should be the caller's responsibility.

if ( symtab_to_fullname ( s ) == NULL )
   /* use symtab->filename */ 
else
   /* use symtab->fullname */

It doesn't really make sense to return the filename and not state that
it is not really the fullname. Also, if the caller tries to access
s->fullname, it will not be successful, because the file simply isn't
there.

>  > RCS file: /cvs/src/src/gdb/source.c,v
>  > retrieving revision 1.49
>  > diff -w -u -r1.49 source.c
>  > --- gdb/source.c	27 Jan 2004 23:19:51 -0000	1.49
>  > +++ gdb/source.c	25 Feb 2004 03:51:45 -0000
>  > @@ -805,30 +805,47 @@
>  >    return 1;
>  >  }
>  >  
>  > -
>  > -/* Open a source file given a symtab S.  Returns a file descriptor or
>  > -   negative number for error.  */
>  > -
>  > +/* This function is capable of finding the absolute path to a
>  > +   source file, and opening it, provided you give it an 
>  > +   OBJFILE and FILENAME. Both the DIRNAME and FULLNAME are only
>  > +   added suggestions on where to find the file. 
>  > +
>  > +   OBJFILE should be the objfile associated with a psymtab or symtab. 
>  > +   FILENAME should be the filename to open.
>  > +   DIRNAME is the compilation directory of a particular source file.
>  > +           Only some debug formats provide this info.
>  > +   FULLNAME can be the last known absolute path to the file in question.
>  > +
>  > +   On Success 
>  > +     A valid file descriptor is returned. ( the return value is positive )
>  > +     FULLNAME is set to the absolute path to the file just opened.
>  > +
>  > +   On Failure
>  > +     A non valid file descriptor is returned. ( the return value is negitive ) 
>  > +     FULLNAME is set to NULL.  */
>  >  int
>  > -open_source_file (struct symtab *s)
>  > +find_and_open_source ( 
>  > +  struct objfile *objfile,	
>  > +  const char *filename,
>  > +  const char *dirname,
>  > +  char **fullname )
>  >  {
> 
> coding standards....

Ok.

>  >    char *path = source_path;
>  >    const char *p;
>  >    int result;
>  > -  char *fullname;
>  >  
>  >    /* Quick way out if we already know its full name */
>  > -  if (s->fullname)
>  > +  if (*fullname)
>  >      {
>  > -      result = open (s->fullname, OPEN_MODE);
>  > +      result = open (*fullname, OPEN_MODE);
>  >        if (result >= 0)
>  >  	return result;
>  >        /* Didn't work -- free old one, try again. */
>  > -      xmfree (s->objfile->md, s->fullname);
>  > -      s->fullname = NULL;
>  > +      xmfree (objfile->md, *fullname);
>  > +      *fullname = NULL;
>  >      }
>  >  
>  > -  if (s->dirname != NULL)
>  > +  if (dirname != NULL)
>  >      {
>  >        /* Replace a path entry of  $cdir  with the compilation directory name */
>  >  #define	cdir_len	5
>  > @@ -841,60 +858,102 @@
>  >  	  int len;
>  >  
>  >  	  path = (char *)
>  > -	    alloca (strlen (source_path) + 1 + strlen (s->dirname) + 1);
>  > +	    alloca (strlen (source_path) + 1 + strlen (dirname) + 1);
>  >  	  len = p - source_path;
>  >  	  strncpy (path, source_path, len);	/* Before $cdir */
>  > -	  strcpy (path + len, s->dirname);	/* new stuff */
>  > +	  strcpy (path + len, dirname);	/* new stuff */
>  >  	  strcat (path + len, source_path + len + cdir_len);	/* After $cdir */
>  >  	}
>  >      }
>  >  
>  > -  result = openp (path, 0, s->filename, OPEN_MODE, 0, &s->fullname);
>  > +  result = openp (path, 0, filename, OPEN_MODE, 0, fullname);
>  >    if (result < 0)
>  >      {
>  >        /* Didn't work.  Try using just the basename. */
>  > -      p = lbasename (s->filename);
>  > -      if (p != s->filename)
>  > -	result = openp (path, 0, p, OPEN_MODE, 0, &s->fullname);
>  > +      p = lbasename (filename);
>  > +      if (p != filename)
>  > +	result = openp (path, 0, p, OPEN_MODE, 0, fullname);
>  >      }
>  >  
>  >    if (result >= 0)
>  >      {
>  > -      fullname = s->fullname;
>  > -      s->fullname = mstrsave (s->objfile->md, s->fullname);
>  > -      xfree (fullname);
>  > +      char *tmp_fullname;
>  > +      tmp_fullname = *fullname;
>  > +      *fullname = mstrsave (objfile->md, *fullname);
>  > +      xfree (tmp_fullname);
>  >      }
>  >    return result;
>  >  }
>  >  
>  > -/* Return the path to the source file associated with symtab.  Returns NULL
>  > -   if no symtab.  */
>  > +/* Open a source file given a symtab S.  Returns a file descriptor or
>  > +   negative number for error.  
>  > +   
>  > +   This function is a convience function to find_and_open_source. */
>  > +
>  > +int
>  > +open_source_file (struct symtab *s)
>  > +{
>  > +    if (!s)
>  > +      return -1;
>  > +
>  > +    return find_and_open_source ( s->objfile, s->filename, s->dirname, &s->fullname );
>  > +}
>  > +
>  > +/* Finds the fullname that a symtab represents.
>  > +
>  > +   If this functions finds the fullname, it will save it in ps->fullname
>  > +   and it will also return the value.
>  >  
>  > +   If this function fails to find the file that this symtab represents,
>  > +   NULL will be returned and ps->fullname will be set to NULL.  */
>  >  char *
>  > -symtab_to_filename (struct symtab *s)
>  > +symtab_to_fullname (struct symtab *s)
>  >  {
>  > -  int fd;
>  > +  int r;
>  >  
>  >    if (!s)
>  >      return NULL;
>  >  
>  > -  /* If we've seen the file before, just return fullname. */
>  > +  /* Don't check s->fullname here, the file could have been 
>  > +     deleted/moved/..., look for it again */
>  > +  r = find_and_open_source ( s->objfile, s->filename, s->dirname, &s->fullname);
>  >  
>  > -  if (s->fullname)
>  > +  if (r)
>  > +  {
>  > +    close (r);
>  >      return s->fullname;
>  > +  }
>  >  
>  > -  /* Try opening the file to setup fullname */
>  > +  return NULL;
>  > +}
>  >  
>  > -  fd = open_source_file (s);
>  > -  if (fd < 0)
>  > -    return s->filename;		/* File not found.  Just use short name */
>  > +/* Finds the fullname that a partial_symtab represents.
>  >  
>  > -  /* Found the file.  Cleanup and return the full name */
>  > +   If this functions finds the fullname, it will save it in ps->fullname
>  > +   and it will also return the value.
>  >  
>  > -  close (fd);
>  > -  return s->fullname;
>  > +   If this function fails to find the file that this partial_symtab represents,
>  > +   NULL will be returned and ps->fullname will be set to NULL.  */
>  > +char *
>  > +psymtab_to_fullname (struct partial_symtab *ps)
>  > +{
>  > +  int r;
>  > +
>  > +  if (!ps)
>  > +    return NULL;
>  > +
>  > +  /* Don't check ps->fullname here, the file could have been
>  > +     deleted/moved/..., look for it again */
>  > +  r = find_and_open_source ( ps->objfile, ps->filename, ps->dirname, &ps->fullname);
>  > +
>  > +  if (r) 
>  > +  {
>  > +    close (r);
>  > +    return ps->fullname;
>  >  }
>  >  
>  > +  return NULL;
>  > +}
>  >  
>  >  /* Create and initialize the table S->line_charpos that records
>  >     the positions of the lines in the source file, which is assumed
> 
> 
> 
>  > Index: gdb/source.h
>  > ===================================================================
>  > Index: gdb/symtab.c
>  > ===================================================================
> 
> These are obvious if the rest goes in.
> 
> 
>  > Index: gdb/symtab.h
>  > ===================================================================
> 
> OK.
> 
> 
>  > Index: gdb/mi/mi-cmd-file.c
>  > ===================================================================
> 
> 
>  > +static const char * const FILENAME = "filename";
>  > +static const char * const FULLNAME = "fullname";
> 
> I don't think these are necessary.

It just unifies the output convention I am using in the
mi-cmd-file unit. What would you prefer to see?

>  >  
>  >  /* Return to the client the absolute path and line number of the 
>  >     current file being executed. */
>  > @@ -39,7 +43,6 @@
>  >    if ( !mi_valid_noargs("mi_cmd_file_list_exec_source_file", argc, argv) )
>  >      error ("mi_cmd_file_list_exec_source_file: Usage: No args");
>  >  
>  > -  
>  >    /* Set the default file and line, also get them */
>  >    set_default_source_symtab_and_line();
>  >    st = get_current_source_symtab_and_line();
>  > @@ -51,17 +54,67 @@
>  >      error ("mi_cmd_file_list_exec_source_file: No symtab");
>  >  
>  >    /* Extract the fullname if it is not known yet */
>  > -  if (st.symtab->fullname == NULL)
>  > -    symtab_to_filename (st.symtab);
>  > -
>  > -  /* We may not be able to open the file (not available). */
>  > -  if (st.symtab->fullname == NULL)
>  > -    error ("mi_cmd_file_list_exec_source_file: File not found");
> 
> Why get rid of the error message?

Ok.

>  > +  symtab_to_fullname (st.symtab);
>  >  
>  >    /* Print to the user the line, filename and fullname */
>  >    ui_out_field_int (uiout, "line", st.line);
>  > -  ui_out_field_string (uiout, "file", st.symtab->filename);
>  > -  ui_out_field_string (uiout, "fullname", st.symtab->fullname);
>  > +  ui_out_field_string (uiout, FILENAME, st.symtab->filename);
>  > +  
>  > +  /* We may not be able to open the file (not available). */
>  > +  if (st.symtab->fullname)
>  > +    ui_out_field_string (uiout, FULLNAME, st.symtab->fullname);
>  > +
> 
> if this test fails shouldn't some warning/error be issued?

I don't know. I am thinking that GDB should just return the absolute
path to all of the source files it can find. If it can not find some,
should it issue a warning? That way the front end could say, "you need
to add a directory to the source search path".

>  > +  return MI_CMD_DONE;
>  > +}
>  > +
>  > +enum mi_cmd_result
>  > +mi_cmd_file_list_exec_source_files(char *command, char **argv, int argc)
>  > +{
>  > +  struct symtab *s;
>  > +  struct partial_symtab *ps;
>  > +  struct objfile *objfile;
>  > +
>  > +  if ( !mi_valid_noargs("mi_cmd_file_list_exec_source_files", argc, argv) )
>  > +    error ("mi_cmd_file_list_exec_source_files: Usage: No args");
>  > +
>  > +  /* Print the table header */
>  > +  ui_out_begin ( uiout, ui_out_type_list, "files");
>  > +
>  > +  /* Look at all of the symtabs */
>  > +  ALL_SYMTABS (objfile, s)
>  > +  {
>  > +    ui_out_begin ( uiout, ui_out_type_tuple, NULL);
>  > +
>  > +    ui_out_field_string (uiout, FILENAME, s->filename);
>  > +
>  > +	/* Extract the fullname if it is not known yet */
>  > +	symtab_to_fullname (s);
>  > +
>  > +	if (s->fullname)
>  > +      ui_out_field_string (uiout, FULLNAME, s->fullname);
>  > +
>  > +    ui_out_end ( uiout, ui_out_type_tuple );
>  > +  }
>  > +
>  > +  /* Look at all of the psymtabs */
>  > +  ALL_PSYMTABS (objfile, ps)
>  > +  {
>  > +    if (!ps->readin) {
> 
> coding standards....

Ok.

>  > +      ui_out_begin ( uiout, ui_out_type_tuple, NULL);
>  > +
>  > +      ui_out_field_string (uiout, FILENAME, ps->filename);
>  > +
>  > +      /* Extract the fullname if it is not known yet */
>  > +	  psymtab_to_fullname (ps);
>  > +
>  > +	  if (ps->fullname) 
>  > +	    ui_out_field_string (uiout, FULLNAME, ps->fullname);
>  > +
>  > +      ui_out_end ( uiout, ui_out_type_tuple );
>  > +    }
>  > +  }
>  > +
>  > +  ui_out_end ( uiout, ui_out_type_list );
>  >  
>  >    return MI_CMD_DONE;
>  >  }
> 
> 
> 
>  > Index: gdb/mi/mi-cmds.c
>  > ===================================================================
> 
>  > Index: gdb/mi/mi-cmds.h
>  > ===================================================================
> 
> these changes are ok.

Great!

>  > Index: gdb/testsuite/gdb.mi/mi-file.exp
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/testsuite/gdb.mi/mi-file.exp,v
>  > retrieving revision 1.1
>  > diff -w -u -r1.1 mi-file.exp
>  > --- gdb/testsuite/gdb.mi/mi-file.exp	2 Apr 2003 22:10:35 -0000	1.1
>  > +++ gdb/testsuite/gdb.mi/mi-file.exp	25 Feb 2004 03:52:36 -0000
>  > @@ -55,7 +55,7 @@
>  >  
>  >      # get the path and absolute path to the current executable
>  >      mi_gdb_test "111-file-list-exec-source-file" \
>  > -	    "111\\\^done,line=\"23\",file=\"${srcfilepath}\",fullname=\"/.*/${srcfile}\"" \
>  > +	    "111\\\^done,line=\"23\",filename=\"${srcfilepath}\",fullname=\"/.*/${srcfile}\"" \
> 
> Wouldn't this break existing MI parsers?

Yes. I figured it could be mi2. Also, for some reason, I thought no one
would be using this function since I wrote it for CGDB, and I haven't
used it yet. I have a larger plan in mind for MI, than just these 2
commands (-file-list-exec-source-file and -file-list-exec-source-files).
I would like to add the fullname to a lot of commands. However, I think
'filename' and 'fullname' should be standardized, so that front end
writers immediatly understand what they are. It is awkard to have 1
function say "file=" and another say "filename=", when those 2 words
mean the same thing. 

However, if this changes isn't acceptable, I can change it back.

>  >                 "request path info of current source file (${srcfile})"
>  >  }
>  >  
>  > Index: gdb/testsuite/gdb.mi/mi2-file.exp
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/testsuite/gdb.mi/mi2-file.exp,v
>  > retrieving revision 1.1
>  > diff -w -u -r1.1 mi2-file.exp
>  > --- gdb/testsuite/gdb.mi/mi2-file.exp	7 Aug 2003 17:47:42 -0000	1.1
>  > +++ gdb/testsuite/gdb.mi/mi2-file.exp	25 Feb 2004 03:52:36 -0000
>  > @@ -47,7 +47,7 @@
>  >  mi_gdb_reinitialize_dir $srcdir/$subdir
>  >  mi_gdb_load ${binfile}
>  >  
>  > -proc test_tbreak_creation_and_listing {} {
>  > +proc test_file_list_exec_source_file {} {
>  >      global srcfile
>  >      global srcdir
>  >      global subdir
>  > @@ -55,11 +55,21 @@
>  >  
>  >      # get the path and absolute path to the current executable
>  >      mi_gdb_test "111-file-list-exec-source-file" \
>  > -	    "111\\\^done,line=\"23\",file=\"${srcfilepath}\",fullname=\"/.*/${srcfile}\"" \
>  > +	    "111\\\^done,line=\"23\",filename=\"${srcfilepath}\",fullname=\"/.*/${srcfile}\"" \
>  >                 "request path info of current source file (${srcfile})"
>  >  }
>  >  
>  > -test_tbreak_creation_and_listing
>  > +proc test_file_list_exec_source_files {} {
>  > +    global srcfile
>  > +
>  > +    # get the path and absolute path to the current executable
>  > +    mi_gdb_test "222-file-list-exec-source-files" \
>  > +	    "222\\\^done,files=\\\[\{filename=\".*/${srcfile}\",fullname=\"/.*/${srcfile}\"\},\{filename=\".*\"\},\{filename=\".*\"\},\{filename=\".*\"\},\{filename=\".*\"\}\\\]" \
> 
>  > +              "Getting a list of source files failed."
>                                                  ^^^^^^^
>                                                   why failed?

OOO, That isn't an error condition, it's just a comment. I see.

Thanks,
Bob Rossi


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