Prefix substitution support for cpp(1)

[Joerg Sonnenberger]

Hi all,
I have a patch in one of my working trees to allow changing the prefix
for __FILE__ based on prefix patterns. This allows building packages
without leaking the full path, so that changes src and obj location
don't alter build.

The primary reason to develop this was to be able to check for
.buildlink references in pkgsrc, which has very similar problems.
It also helps if diagnostic messages actually refer to the files in
/usr/pkg/include.

Would this be considered generally useful? In that case I will clean up
the patch and prepare some documentation for it.

Joerg

[Alan Barrett]

On Sun, 25 Oct 2009, Joerg Sonnenberger wrote:
> I have a patch in one of my working trees to allow changing the prefix
> for __FILE__ based on prefix patterns. This allows building packages
> without leaking the full path, so that changes src and obj location
> don't alter build.

That would be very useful!

--apb (Alan Barrett)

[David Young]

On Sun, Oct 25, 2009 at 05:24:19PM +0100, Joerg Sonnenberger wrote:
> Hi all,
> I have a patch in one of my working trees to allow changing the prefix
> for __FILE__ based on prefix patterns. This allows building packages
> without leaking the full path, so that changes src and obj location
> don't alter build.
*snip snip*
> Would this be considered generally useful? In that case I will clean up
> the patch and prepare some documentation for it.

Very useful!  Do you plan to tackle the debugger symbols, too?

Dave

--
David Young             OJC Technologies
dyoung@...      Urbana, IL * [phone]

[Joerg Sonnenberger]

On Sun, Oct 25, 2009 at 03:04:12PM -0500, David Young wrote:

> On Sun, Oct 25, 2009 at 05:24:19PM +0100, Joerg Sonnenberger wrote:
> > Hi all,
> > I have a patch in one of my working trees to allow changing the prefix
> > for __FILE__ based on prefix patterns. This allows building packages
> > without leaking the full path, so that changes src and obj location
> > don't alter build.
> *snip snip*
> > Would this be considered generally useful? In that case I will clean up
> > the patch and prepare some documentation for it.
>
> Very useful!  Do you plan to tackle the debugger symbols, too?

I'm not sure if that can be expressed easily. It also something that
would be better served by a separate program to transform debug
information. This strings are not hard-coded into the programs after
all. So: not as part of this change.

Joerg

[Joerg Sonnenberger]

On Sun, Oct 25, 2009 at 05:24:19PM +0100, Joerg Sonnenberger wrote:
> I have a patch in one of my working trees to allow changing the prefix
> for __FILE__ based on prefix patterns. This allows building packages
> without leaking the full path, so that changes src and obj location
> don't alter build.

Attached is a cleaned up patch. Suggestions for a better option name?

Joerg

Index: dist/gcc4/gcc/c-opts.c
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/c-opts.c,v
retrieving revision 1.4
diff -u -p -r1.4 c-opts.c
--- dist/gcc4/gcc/c-opts.c 21 Jun 2007 04:02:00 -0000 1.4
+++ dist/gcc4/gcc/c-opts.c 30 Oct 2009 19:44:08 -0000
@@ -163,6 +163,7 @@ c_common_missing_argument (const char *o
     case OPT_isysroot:
     case OPT_isystem:
     case OPT_iquote:
+    case OPT_iunwrap:
       error ("missing path after %qs", opt);
       break;
 
@@ -841,6 +842,10 @@ c_common_handle_option (size_t scode, co
       add_path (xstrdup (arg), SYSTEM, 0, true);
       break;
 
+    case OPT_iunwrap:
+      add_cpp_unwrap_path (arg);
+      break;
+
     case OPT_iwithprefix:
       add_prefixed_path (arg, SYSTEM);
       break;
Index: dist/gcc4/gcc/c.opt
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/c.opt,v
retrieving revision 1.3
diff -u -p -r1.3 c.opt
--- dist/gcc4/gcc/c.opt 21 Jun 2007 04:02:00 -0000 1.3
+++ dist/gcc4/gcc/c.opt 30 Oct 2009 19:37:36 -0000
@@ -783,6 +783,10 @@ iquote
 C ObjC C++ ObjC++ Joined Separate
 -iquote <dir> Add <dir> to the end of the quote include path
 
+iunwrap
+C ObjC C++ ObjC++ Joined Separate
+-iunwrap <src:dst> Convert <src> to <dst> if it occurs as prefix in __FILE__.
+
 iwithprefix
 C ObjC C++ ObjC++ Joined Separate
 -iwithprefix <dir> Add <dir> to the end of the system include path
Index: dist/gcc4/gcc/gcc.h
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/gcc.h,v
retrieving revision 1.2
diff -u -p -r1.2 gcc.h
--- dist/gcc4/gcc/gcc.h 12 May 2006 00:33:26 -0000 1.2
+++ dist/gcc4/gcc/gcc.h 30 Oct 2009 19:37:59 -0000
@@ -50,7 +50,7 @@ struct spec_function
   || !strcmp (STR, "idirafter") || !strcmp (STR, "iprefix") \
   || !strcmp (STR, "iwithprefix") || !strcmp (STR, "iwithprefixbefore") \
   || !strcmp (STR, "iquote") || !strcmp (STR, "isystem") \
-  || !strcmp (STR, "isysroot") \
+  || !strcmp (STR, "-iunwrap") || !strcmp (STR, "isysroot") \
   || !strcmp (STR, "-param") || !strcmp (STR, "specs") \
   || !strcmp (STR, "MF") || !strcmp (STR, "MT") || !strcmp (STR, "MQ"))
 
Index: dist/gcc4/gcc/doc/cpp.texi
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/doc/cpp.texi,v
retrieving revision 1.4
diff -u -p -r1.4 cpp.texi
--- dist/gcc4/gcc/doc/cpp.texi 21 Jun 2007 04:02:02 -0000 1.4
+++ dist/gcc4/gcc/doc/cpp.texi 30 Oct 2009 21:58:20 -0000
@@ -4196,6 +4196,7 @@ without notice.
 @c man begin SYNOPSIS
 cpp [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
     [@option{-I}@var{dir}@dots{}] [@option{-iquote}@var{dir}@dots{}]
+    [@option{-iunwrap}@var{src}:@var{dst}]
     [@option{-W}@var{warn}@dots{}]
     [@option{-M}|@option{-MM}] [@option{-MG}] [@option{-MF} @var{filename}]
     [@option{-MP}] [@option{-MQ} @var{target}@dots{}]
Index: dist/gcc4/gcc/doc/cppopts.texi
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/doc/cppopts.texi,v
retrieving revision 1.2
diff -u -p -r1.2 cppopts.texi
--- dist/gcc4/gcc/doc/cppopts.texi 12 May 2006 00:33:26 -0000 1.2
+++ dist/gcc4/gcc/doc/cppopts.texi 30 Oct 2009 21:58:32 -0000
@@ -513,6 +513,12 @@ Search @var{dir} only for header files r
 @xref{Search Path}.
 @end ifset
 
+@item -iunwrap @var{src}:@var{dst}
+@opindex iunwrap
+Replace the prefix @var{src} in __FILE__ with @var{dst} at expansion time.
+This option can be specified more than once.  Processing stops at the first
+match.
+
 @item -fdollars-in-identifiers
 @opindex fdollars-in-identifiers
 @anchor{fdollars-in-identifiers}
Index: dist/gcc4/gcc/doc/invoke.texi
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/gcc/doc/invoke.texi,v
retrieving revision 1.8
diff -u -p -r1.8 invoke.texi
--- dist/gcc4/gcc/doc/invoke.texi 29 Apr 2009 23:53:28 -0000 1.8
+++ dist/gcc4/gcc/doc/invoke.texi 30 Oct 2009 21:57:20 -0000
@@ -376,7 +376,8 @@ Objective-C and Objective-C++ Dialects}.
 
 @item Directory Options
 @xref{Directory Options,,Options for Directory Search}.
-@gccoptlist{-B@var{prefix}  -I@var{dir}  -iquote@var{dir}  -L@var{dir}
+@gccoptlist{-B@var{prefix}  -I@var{dir}  -iquote@var{dir}  
+-iunwrap@var{src}:@var{dst}  -L@var{dir}
 -specs=@var{file}  -I- --sysroot=@var{dir}}
 
 @item Target Options
@@ -6506,6 +6507,12 @@ be searched for header files only for th
 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
 otherwise just like @option{-I}.
 
+@item -iunwrap @var{src}:@var{dst}
+@opindex iunwrap
+Replace the prefix @var{src} in __FILE__ with @var{dst} at expansion time.
+This option can be specified more than once.  Processing stops at the first
+match.
+
 @item -L@var{dir}
 @opindex L
 Add directory @var{dir} to the list of directories to be searched
Index: dist/gcc4/libcpp/macro.c
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/libcpp/macro.c,v
retrieving revision 1.2
diff -u -p -r1.2 macro.c
--- dist/gcc4/libcpp/macro.c 30 May 2008 15:12:24 -0000 1.2
+++ dist/gcc4/libcpp/macro.c 30 Oct 2009 21:52:49 -0000
@@ -109,6 +109,61 @@ static const char * const monthnames[] =
   "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
 };
 
+static size_t unwrap_pairs;
+static char **unwrap_src;
+static char **unwrap_dst;
+
+void
+add_cpp_unwrap_path (const char *arg)
+{
+ const char *arg_dst;
+ size_t len;
+
+ arg_dst = strchr(arg, ':');
+ if (arg_dst == NULL) {
+ fprintf(stderr, "Invalid argument for -iunwrap");
+ exit(1);
+ }
+ len = arg_dst - arg;
+ ++arg_dst;
+
+ unwrap_src = xrealloc(unwrap_src, sizeof(char *) * (unwrap_pairs + 1));
+ unwrap_dst = xrealloc(unwrap_dst, sizeof(char *) * (unwrap_pairs + 1));
+
+ unwrap_src[unwrap_pairs] = xmalloc(len + 1);
+ memcpy(unwrap_src[unwrap_pairs], arg, len);
+ unwrap_src[unwrap_pairs][len] = '\0';
+ unwrap_dst[unwrap_pairs] = xstrdup(arg_dst);
+ ++unwrap_pairs;
+}
+
+static const char *
+cpp_unwrap_file (const char *arg, char **tmp_name)
+{
+ char *result;
+ size_t i, len;
+
+ for (i = 0; i < unwrap_pairs; ++i) {
+ len = strlen (unwrap_src[i]);
+ if (strncmp (unwrap_src[i], arg, len))
+ continue;
+ if (arg[len] == '\0')
+ return xstrdup (unwrap_dst[i]);
+ if (arg[len] != '/')
+ continue;
+ arg += len;
+ len = strlen (unwrap_dst[i]);
+ result = xmalloc (len + strlen (arg) + 1);
+ memcpy(result, unwrap_dst[i], len);
+ strcpy(result + len, arg);
+ *tmp_name = result;
+
+ return result;
+ }
+
+ return arg;
+}
+
 /* Helper function for builtin_macro.  Returns the text generated by
    a builtin macro. */
 const uchar *
@@ -130,6 +185,7 @@ _cpp_builtin_macro_text (cpp_reader *pfi
       {
  unsigned int len;
  const char *name;
+ char *tmp_name;
  uchar *buf;
  map = linemap_lookup (pfile->line_table, pfile->line_table->highest_line);
 
@@ -137,12 +193,14 @@ _cpp_builtin_macro_text (cpp_reader *pfi
   while (! MAIN_FILE_P (map))
     map = INCLUDED_FROM (pfile->line_table, map);
 
- name = map->to_file;
+ tmp_name = NULL;
+ name = cpp_unwrap_file (map->to_file, &tmp_name);
  len = strlen (name);
  buf = _cpp_unaligned_alloc (pfile, len * 2 + 3);
  result = buf;
  *buf = '"';
  buf = cpp_quote_string (buf + 1, (const unsigned char *) name, len);
+ free (tmp_name);
  *buf++ = '"';
  *buf = '\0';
       }
Index: dist/gcc4/libcpp/include/cpplib.h
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/dist/gcc4/libcpp/include/cpplib.h,v
retrieving revision 1.1.1.1
diff -u -p -r1.1.1.1 cpplib.h
--- dist/gcc4/libcpp/include/cpplib.h 20 Apr 2006 09:55:51 -0000 1.1.1.1
+++ dist/gcc4/libcpp/include/cpplib.h 30 Oct 2009 19:43:35 -0000
@@ -623,6 +623,9 @@ extern void cpp_set_lang (cpp_reader *,
 /* Set the include paths.  */
 extern void cpp_set_include_chains (cpp_reader *, cpp_dir *, cpp_dir *, int);
 
+/* Provide src:dst pair for __FILE__ unwrapping.  */
+extern void add_cpp_unwrap_path (const char *);
+
 /* Call these to get pointers to the options, callback, and deps
    structures for a given reader.  These pointers are good until you
    call cpp_finish on that reader.  You can either edit the callbacks
Index: usr.bin/gcc4/cpp/cpp.1
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/usr.bin/gcc4/cpp/cpp.1,v
retrieving revision 1.2
diff -u -p -r1.2 cpp.1
--- usr.bin/gcc4/cpp/cpp.1 30 Apr 2009 00:45:47 -0000 1.2
+++ usr.bin/gcc4/cpp/cpp.1 30 Oct 2009 22:00:26 -0000
@@ -11,6 +11,7 @@
 .Op Fl U Ns Ar macro
 .Op Fl I Ns Ar dir...
 .Op Fl iquote Ns Ar dir...
+.Op Fl iunwrap Ns Ar src Ns : Ns Ar dst
 .Op Fl W Ns Ar warn...
 .Op Fl M | Fl MM
 .Op Fl MG
@@ -691,6 +692,16 @@ before all directories specified by
 .Fl I
 and before the standard system directories.
 .Pp
+.It Fl iunwrap Ns Ar src Ns : Ns Ar dst
+Replace the prefix
+.Ar src
+in
+.Va __FILE__
+with
+.Ar dst
+at expansion time.
+This option can be specified more than once.
+Processing stops at the first match.
 .It Fl fdollars-in-identifiers
 @anchor{fdollars-in-identifiers} Accept
 .Sy $
Index: usr.bin/gcc4/gcc/gcc.1
===================================================================
RCS file: /home/joerg/repo/netbsd/src/gnu/usr.bin/gcc4/gcc/gcc.1,v
retrieving revision 1.3
diff -u -p -r1.3 gcc.1
--- usr.bin/gcc4/gcc/gcc.1 30 Apr 2009 00:30:56 -0000 1.3
+++ usr.bin/gcc4/gcc/gcc.1 30 Oct 2009 22:01:16 -0000
@@ -183,6 +183,7 @@ Explanations are in the following sectio
 .Fl B Ns Ar prefix
 .Fl I Ns Ar dir
 .Fl iquote Ns Ar dir
+.Fl iunwrap Ns Ar src Ns : Ns Ar dst
 .Fl L Ns Ar dir
 .Fl specs= Ns Ar file
 .Fl I- Fl -sysroot= Ns Ar dir
@@ -6633,6 +6634,16 @@ before all directories specified by
 .Fl I
 and before the standard system directories.
 .Pp
+.It Fl iunwrap Ns Ar src Ns : Ns Ar dst
+Replace the prefix
+.Ar src
+in
+.Va __FILE__
+with
+.Ar dst
+at expansion time.
+This option can be specified more than once.
+Processing stops at the first match.
 .It Fl fdollars-in-identifiers
 @anchor{fdollars-in-identifiers} Accept
 .Sy $

[Iain Hibbert]

On Sat, 31 Oct 2009, Joerg Sonnenberger wrote:

> Suggestions for a better option name?

"rewrap" (remap?) is better than "unwrap"


iain

[Joerg Sonnenberger]

On Fri, Oct 30, 2009 at 11:56:22PM +0000, Iain Hibbert wrote:
> On Sat, 31 Oct 2009, Joerg Sonnenberger wrote:
>
> > Suggestions for a better option name?
>
> "rewrap" (remap?) is better than "unwrap"

remap in that case. Any other opinions?

Joerg