[patch] Bitlbee 1.2.1 breaks glib 2.4 compatibility, uses g_file_test() from glib 2.8

[robert@…]

I'm the Fedora EPEL maintainer of bitlbee and RHEL 4 (CentOS 4) has glib 2.4.7, which doesn't work with bitlbee 1.2.1 any longer:

glib/gstdio.h: No such file or directory
storage_xml.c: In function `xml_init':
storage_xml.c:248: warning: implicit declaration of function `g_access'

The attached patch solves the problem for me. As far as I got, g_access() is only required for Windows compatibility or porting things. So if no g_access() exists, access() can be used further on in most cases AKA Linux. Was broken with Changeset #385 by Jelmer Vernooij.

[robert@…]

bitlbee-1.2.1-g_access.patch

[Wilmer van der Gaast <wilmer@…>]

#429: [patch] Bitlbee 1.2.1 breaks glib 2.4 compatibility, uses g_file_test()
from glib 2.8

storage_xml.c: In function `xml_init':
storage_xml.c:248: warning: implicit declaration of function `g_access'

So is g_access() the problem or g_file_test()? Jelmer assured me that
g_file_test() works in GLib 2.4, and indeed I don't see any clear sign
that it doesn't, it's only g_access() that sucks. I asked Jelmer what's
the difference between the two. I expect that just using g_file_test()
everywhere should solve this problem.

[robert@…]

Sorry, my failure. It's only g_access() which makes the problem with glib 2.4. I didn't have a correct look to it.

* Compiling storage_xml.c
storage_xml.c:31:25:

glib/gstdio.h: No such file or directory
storage_xml.c: In function `xml_init':
storage_xml.c:248: warning: implicit declaration of function `g_access'
make: *** [storage_xml.o] Error 1

[robert@…]

bitlbee-1.2.1-g_access.patch

[robert@…]

​http://library.gnome.org/devel/glib/unstable/glib-File-Utilities.html#GFileTest - there's no check for testing writeable with g_file_test() or am I wrong?

[wilmer]

Yup, Jelmer just told me that too. I submitted his fix, and altered the code to just use g_access() everywhere (I find using both that and g_file_test() a little bit messy). Jelmer's fix adds a #define g_access access, which should work well enough.

changeset:devel,408

[robert@…]

We are still missing the conditional include, because the file does not exist in old versions (see comments above). And shouldn't we define F_OK again as in the past, if it is maybe not defined?! But in the past, the F_OK was defined in another file, see the link above. Note, that my diff is against Bitlbee 1.2.1, not against current changesets. But should be easily understandable as well.

--- bitlbee-1.2.1/storage_xml.c           2008-06-10 01:09:06.000000000 +0200
+++ bitlbee-1.2.1/storage_xml.c.gaccess   2008-06-30 22:45:33.000000000 +0200
@@ -28,8 +28,18 @@
 #include "base64.h"
 #include "arc.h"
 #include "md5.h"
+
+#if !GLIB_CHECK_VERSION(2,8,0)
 #include <glib/gstdio.h>

+/* GLib < 2.8.0 doesn't have g_access, so just use the system access(). */
+#define g_access access
+
+#ifndef F_OK
+#define F_OK 0
+#endif
+#endif
+
 typedef enum
 {
        XML_PASS_CHECK_ONLY = -1,

[robert@…]

Once upstream currently seems to be unable to provide a fully working solution, the following (working) patch went into Fedora and EPEL:

​http://cvs.fedoraproject.org/viewcvs/*checkout*/devel/bitlbee/bitlbee-1.2.1-gaccess.patch

[wilmer]

This really should be fixed now.

changeset:devel,418.

[robert@…]

No, no, no, Wilmer, no. This is still not fixed, unluckily. It still fails in RHEL 4.x once more. But all good things are three, right? So you've got a third try now - maybe fixed for Bitlbee 1.2.3? You know, I'm just kidding... ;-)

Once this is covered by upstream, the following (working) patch went into Fedora and EPEL:

​http://cvs.fedoraproject.org/viewcvs/*checkout*/devel/bitlbee/bitlbee-1.2.2-gaccess.patch

[robert@…]

Fixed with Bitlbee 1.2.3 now - can somebody please look for the correct changeset and add it here and close then? Looks like I'm somehow not allowed to close.

[wilmer]

changeset:devel,443 and changeset:devel,tag:1.2.3 (I doubt if Trac supports that syntax, though.)

Yeah, sorry, those controls are closed since spammers liked to mess with them.