xgettext: problems with PHP heredocs

[Gaëtan Frenoy]

Package: gettext-base
Version: 0.14.4-2

Hello,

Here is a patch to an bug I discovered in xgettext.
Note that this problem was already reported a while ago
in gnu.utils.bug
(see
http://groups.google.be/group/gnu.utils.bug/tree/browse_frm/thread/94bed260e
b3dde71/e8c679cb94fb3b8a)

But so far, I did not find any fix.

First let's describe the problem.  Say you have the
following PHP source file.

-------- bug_heredoc.php -------------------------- start -
<?php

$foo = _("Bar");
echo <<<_END_
[...]
<script language="javascript">
[...]
</script>
_END_;
$foo2 = _("Bar2");

?>
-------- bug_heredoc.php ---------------------------- end -

Now, run "xgettext" to extract marked strings :

-- xgettext call ---------------------------------- start -
$ xgettext -L PHP --omit-header heredoc_bug.php -o -
#: heredoc_bug.php:3
msgid "Bar"
msgstr ""
-- xgettext call ------------------------------------ end -

Surprisingly, "Bar2" is not reported nor any of marked
strings located after the end of PHP heredoc section.

If you are not familiar with PHP, here are some words
about Heredoc syntax:
http://www.php.net/manual/en/language.types.string.php#language.types.string
.syntax.heredoc

I am running the software version on a Debian stable
("Sarge") system :

-- xgettext version ------------------------------ start -
$ xgettext --version
xgettext (GNU gettext-tools) 0.14.4
(...)

$ apt-cache policy gettext
gettext:
  Installé : 0.14.4-2
  Candidat : 0.14.4-2
 Table de version :
 *** 0.14.4-2 0
        500 ftp://ftp.belnet.be sarge/main Packages
        100 /var/lib/dpkg/status
-- xgettext version -------------------------------- end -

However, it looks like the problem is not yet fixed in
0.14.5.


By digging into the code, I found the following fix 
for gettext-tools/src/x-php.c :

-- x-php.c patch --------------------------------- start -
$ diff -abu x-php.orig.c x-php.c
--- x-php.orig.c        2003-12-30 12:30:01.000000000 +0100
+++ x-php.c     2006-05-04 19:14:43.434424200 +0200
@@ -1087,12 +1087,18 @@
                  {
                    int bufidx = 0;

+                   /* Skip blank lines before processing
+                    * possible label */
+                   do
+                     c = phase1_getc ();
+                    while (c != EOF && (c == '\n' || c == '\r'));
+
                    while (bufidx < bufpos)
                      {
                        c = phase1_getc ();
                        if (c == EOF)
                          break;
-                       if (c != buffer[bufidx])
+                       if (c != buffer[bufidx++])
                          {
                            phase1_ungetc (c);
                            break;
-- x-php.c patch ----------------------------------- end -


And now xgettext behaves as expected.

Hope this helps.

-- 
Gaëtan Frenoy (gaetan -à- frenoy -point- net)