Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Frederic RISS <frederic.riss@st.com>
To: Jim Blandy <jimb@red-bean.com>
Cc: Daniel Jacobowitz <drow@false.org>,
	        GDB Patches <gdb-patches@sourceware.org>
Subject: Re: [RFC] Don't lose compilation directory in Dwarf2 line-tables
Date: Fri, 21 Apr 2006 08:35:00 -0000	[thread overview]
Message-ID: <1145608477.14807.1018.camel@crx549.cro.st.com> (raw)
In-Reply-To: <8f2776cb0604201809l68067aey47fa372a2468c8ea@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1257 bytes --]

On Thu, 2006-04-20 at 18:09 -0700, Jim Blandy wrote:
> So, Frederic, your last patch doesn't introduce any new problems that
> I can see, and it does solve the problem you set out to solve
> originally (losing comp_dir), but if you're willing go around one more
> time, we can try to add the info consistently:

OK, here we go.

> - Leave the comparison loop alone, as in your last patch.
> - If dwarf2_start_subfile does have to start a subfile itself, always
> pass comp_dir as start_subfile's second argument, whether it's NULL or
> not (because this is what we do when calling start_symtab), and
> concatenate dirname, if it's non-null, with filename to get
> start_subfile's first argument.  I think this means that 'fullname'
> always gets used, so you can hoist that computation and its xfree out
> of the 'if'.

But then, the loop in dwarf2_start_subfile doesn't serve any purpose,
because the loop at the beginning of start_subfile proper will do the
same work. Or maybe I'm missing something?

> Either way, this definitely needs a comment.  If you'd like to write
> up one yourself, great; if not, that's fine; I'll put one in after
> your patch goes in.

Attached is a new patch that adds comments and removes the superfluous
loop. How's that?


[-- Attachment #2: dwarf2-keep-compdir.patch --]
[-- Type: text/x-patch, Size: 4620 bytes --]

2006-04-21  Frederic Riss  <frederic.riss@st.com>

	* dwarf2read.c (dwarf2_start_subfile): Change prototype to accept
	compilation directory as last argument. 
	Always pass comp_dir as second argument to start_subfile and prepend
	dirname to the filename when necessary. 
	Remove now superfluous search for pre-existing subfile.
	(dwarf_decode_lines): Pass the compilation directory to
	dwarf2_start_subfile.
	


--- dwarf2read.c.orig	2006-04-13 11:17:40.000000000 +0200
+++ dwarf2read.c	2006-04-21 10:02:08.000000000 +0200
@@ -846,7 +846,7 @@
 static void dwarf_decode_lines (struct line_header *, char *, bfd *,
 				struct dwarf2_cu *, struct partial_symtab *);
 
-static void dwarf2_start_subfile (char *, char *);
+static void dwarf2_start_subfile (char *, char *, char *);
 
 static struct symbol *new_symbol (struct die_info *, struct type *,
 				  struct dwarf2_cu *);
@@ -6529,13 +6529,12 @@
 	     directory and file name numbers in the statement program
 	     are 1-based.  */
           struct file_entry *fe = &lh->file_names[file - 1];
-          char *dir;
+          char *dir = NULL;
 
           if (fe->dir_index)
             dir = lh->include_dirs[fe->dir_index - 1];
-          else
-            dir = comp_dir;
-	  dwarf2_start_subfile (fe->name, dir);
+
+	  dwarf2_start_subfile (fe->name, dir, comp_dir);
 	}
 
       /* Decode the table.  */
@@ -6627,17 +6626,16 @@
                    0-based, but the directory and file name numbers in
                    the statement program are 1-based.  */
                 struct file_entry *fe;
-                char *dir;
+                char *dir = NULL;
 
                 file = read_unsigned_leb128 (abfd, line_ptr, &bytes_read);
                 line_ptr += bytes_read;
                 fe = &lh->file_names[file - 1];
                 if (fe->dir_index)
                   dir = lh->include_dirs[fe->dir_index - 1];
-                else
-                  dir = comp_dir;
+
                 if (!decode_for_pst_p)
-                  dwarf2_start_subfile (fe->name, dir);
+                  dwarf2_start_subfile (fe->name, dir, comp_dir);
               }
 	      break;
 	    case DW_LNS_set_column:
@@ -6717,7 +6715,8 @@
 
 /* Start a subfile for DWARF.  FILENAME is the name of the file and
    DIRNAME the name of the source directory which contains FILENAME
-   or NULL if not known.
+   or NULL if not known.  COMP_DIR is the compilation directory for the
+   linetable's compilation unit or NULL if not known.
    This routine tries to keep line numbers from identical absolute and
    relative file names in a common subfile.
 
@@ -6733,31 +6732,35 @@
    files.files[1].dir:  /srcdir
 
    The line number information for list0.c has to end up in a single
-   subfile, so that `break /srcdir/list0.c:1' works as expected.  */
-
-static void
-dwarf2_start_subfile (char *filename, char *dirname)
-{
-  /* If the filename isn't absolute, try to match an existing subfile
-     with the full pathname.  */
+   subfile, so that `break /srcdir/list0.c:1' works as expected.
+   start_subfile will ensure that this happens provided that we pass the
+   concatenation of files.files[1].dir and files.files[1].name as the
+   subfile's name.  */
+
+static void
+dwarf2_start_subfile (char *filename, char *dirname, char *comp_dir)
+{
+  char *fullname;
+
+  /* While reading the DIEs, we call start_symtab(DW_AT_name, DW_AT_comp_dir).
+     `start_symtab' will always pass the contents of DW_AT_comp_dir as
+     second argument to start_subfile.  To be consistent, we do the
+     same here.  In order not to lose the line information directory,
+     we concatenate it to the filename when it makes sense.
+     Note that the Dwarf3 standard says (speaking of filenames in line
+     information): ``The directory index is ignored for file names
+     that represent full path names''.  Thus ignoring dirname in the
+     `else' branch below isn't an issue.  */
 
   if (!IS_ABSOLUTE_PATH (filename) && dirname != NULL)
-    {
-      struct subfile *subfile;
-      char *fullname = concat (dirname, "/", filename, (char *)NULL);
+    fullname = concat (dirname, SLASH_STRING, filename, (char *)NULL);
+  else
+    fullname = filename;
 
-      for (subfile = subfiles; subfile; subfile = subfile->next)
-	{
-	  if (FILENAME_CMP (subfile->name, fullname) == 0)
-	    {
-	      current_subfile = subfile;
-	      xfree (fullname);
-	      return;
-	    }
-	}
-      xfree (fullname);
-    }
-  start_subfile (filename, dirname);
+  start_subfile (fullname, comp_dir);
+
+  if (fullname != filename)
+    xfree (fullname);
 }
 
 static void

  reply	other threads:[~2006-04-21  8:35 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-04-13 11:25 [RFC] Don't loose " Frederic RISS
2006-04-13 17:49 ` Jim Blandy
2006-04-14  7:32   ` [RFC] Don't lose " Frederic RISS
2006-04-14  7:41     ` Jim Blandy
2006-04-14  8:23       ` Frederic RISS
2006-04-14 14:04         ` Daniel Jacobowitz
2006-04-14 14:51           ` Frederic RISS
2006-04-18 12:32             ` Frederic RISS
2006-04-18 13:04               ` Daniel Jacobowitz
2006-04-18 14:00                 ` Frederic RISS
2006-04-20 16:27                   ` Daniel Jacobowitz
2006-04-21  7:14                     ` Frederic RISS
2006-04-21  1:10                   ` Jim Blandy
2006-04-21  8:35                     ` Frederic RISS [this message]
2006-04-21 16:46                       ` Jim Blandy
2006-04-21 17:00                         ` Frederic RISS
2006-04-24 12:49                         ` Andrew STUBBS
2006-04-14  8:44     ` Eli Zaretskii
2006-04-14 11:44       ` Frederic RISS
2006-04-14 13:08         ` Eli Zaretskii
2006-04-14 13:42           ` Frederic RISS
2006-04-14 14:21             ` Eli Zaretskii
2006-04-14 14:58               ` Daniel Jacobowitz
2006-04-14 15:03                 ` Eli Zaretskii
2006-04-14 18:39               ` Frédéric Riss
2006-04-13 19:58 ` [RFC] Don't loose " Jason Molenda

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1145608477.14807.1018.camel@crx549.cro.st.com \
    --to=frederic.riss@st.com \
    --cc=drow@false.org \
    --cc=gdb-patches@sourceware.org \
    --cc=jimb@red-bean.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox