[Geany] [PATCH] link vte in dynamically instead of using g_module_open

[Mark Knoop]

On 24/01/07, Josef Whiter <josef at toxicpanda.com> wrote:
> Enrico Tröger wrote:
> > On Wed, 24 Jan 2007 12:26:34 -0500, Josef Whiter <josef at toxicpanda.com>
> > wrote:
> >
> > Hi,
> >
> >
> >> Geany doesn't find the vte lib in fedora when you have just the vte
> >> package installed, but it works fine when you have vte-devel
> >> installed. This is because the normal vte package is something like
> >> libvte.so.0.9.4 or something like that, and geany doesn't pick it up,
> >>
> > You had a look at vte.c, so you should know that Geany looks for some
> > so-names as fallback.
> > On Debian, I have the two files
> > libvte.so.4 and libvte.so.4.5.1 where libvte.so.4 is just a symlink to
> > the other one. Why doesn't Fedora does it in that way?
> >
> >
> >> however vte-devel installs libvte.so which geany does pick up, and
> >> since i need vte-devel to build the geany package for fedora I didn't
> >>
> > Why do you need the vte-devel package to build Geany? Geany should
> > build independent of the existence of any VTE files, also with enabled
> > VTE support.
> > Just tested with a fresh svn checkout, ran autogen.sh and make. I don't
> > have the vte-dev package installed and all went as expected. The VTE
> > support was compiled in and I could use it.
> >
> >
> >> notice this, but if you don't have vte-devel installed geany
> >> complains about not being able to load the terminal.  Since I can't
> >> require a -devel package for a normal package I have to make geany
> >> link vte in during compile instead of doing g_module_open.  I just
> >> finished this patch this morning and tested it and it works well.
> >> Will you guys include this?  Or is there a specific reason you use
> >> g_module_open and g_module_symbol?  It looks like you were putting it
> >>
> > Yes, the reason is to not depend on libvte. The idea was to laod the
> > module if it is found and use it. If it can't be found, just let out the
> > VTE component.
> >
> >
> >> in the configure script at one point, but just commented it out.
> >>
> > Yes, it was a try to detect the lib with the autotools but it is not
> > possible(at least I don't know how) without having installed any dev
> > packages.
> >
> > I won't apply your patch because I don't want to add the general
> > dependency on libvte. We can add libvte.so.0.9.4 to the list of
> > searched modules. But are you sure that on Fedora there is no symlink
> > to this file without the patch level?
> >
> > Nick, how it is handled on your system?
> >
> >
> Hmm, probably should have asked before doing :).  It looks like our vte
> package does make a link for /usr/lib/libvte.so.9, but for some reason
> geany was complaining about not being able to find it, even though its
> in the list of backup modules to look for.  I will look into why this
> isn't behaving properly.  Thanks much,

The problem seems to be in vte.c - the if else construct looking for
libvte versions will never go past the first check. Works with this
patch:

--- vte.c       2007-01-25 10:18:32.000000000 +0000
+++ vte.c.old   2007-01-25 10:18:29.000000000 +0000
@@ -137,8 +137,8 @@
                module = g_module_open("libvte.so", G_MODULE_BIND_LAZY);
                // try to fallback to different versions of libvte.so.x
                if (module == NULL) module =
g_module_open("libvte.so.4", G_MODULE_BIND_LAZY);
-               if (module == NULL) module =
g_module_open("libvte.so.8", G_MODULE_BIND_LAZY);
-               if (module == NULL) module =
g_module_open("libvte.so.9", G_MODULE_BIND_LAZY);
+               else if (module == NULL) module =
g_module_open("libvte.so.8", G_MODULE_BIND_LAZY);
+               else if (module == NULL) module =
g_module_open("libvte.so.9", G_MODULE_BIND_LAZY);
        }

        if (module == NULL)


-- 
Mark Knoop