update midori to 0.2.6

[elelay (Eric Le Lay)]

The latest release of midori looks good. Could it be updated ?

I had to patch it a bit to have extensions working.

[ryandesign (Ryan Carsten Schmidt)]

Instead of hardcoding "0.2" in the master_sites, you should define and use a branch variable; see PortfileRecipes.

The no_x11 variant you added is insufficient because all it does is remove a dependency.

[elelay (Eric Le Lay)]

Replying to ryandesign@…:

Instead of hardcoding "0.2" in the master_sites, you should define and use a branch variable; see PortfileRecipes.

thanks for your input, and for the link to PortfileRecipes : lots of interesting stuff.

The no_x11 variant you added is insufficient because all it does is remove a dependency.

Do you mean that if libnotify happens to be installed, it will be picked even if it doesn't work ? fixing this...

[elelay (Eric Le Lay)]

update midori to 0.2.6 + don't use libnotify with no_x11

[afb@…]

The patch-midori-main-c.diff change is a hack, it should build bundles rather than dynamiclibs.

Something like the below, unless the silly build system used has been improved since last time:

reinplace "s/shlib_LINKFLAGS = \\\['-dynamiclib'\\\]/shlib_LINKFLAGS = \['-bundle'\]/" \
          ${worksrcpath}/_build_/c4che/default.cache.py
reinplace "s/shlib_PATTERN = 'lib%s.dylib'/shlib_PATTERN = 'lib%s.so'/" \
          ${worksrcpath}/_build_/c4che/default.cache.py

[elelay (Eric Le Lay)]

err, well I don't know the relative merits of bundles vs dynamic libraries. What I found rather stupid was the definition of G_MODULE_SUFFIX to "so" in the first place. We are on a mac here : libraries are ending with ".dylib".

But you are 100% right : it's a quick hack to just "have it working". Please use any other means you like better. It's just that I stumbled upon midori when trying to update webkit-gtk as a dependency for gPodder. I updated it as a testbed for my webkit-gtk build.

[afb@…]

Replying to elelay@…:

We are on a mac here : libraries are ending with ".dylib".

Loadable bundles are not dynamic libraries, so they end with ".bundle" or ".so".

[elelay (Eric Le Lay)]

Meanwhile, midori 0.2.7 has been released.

I don't know if you have taken a close look at midori/main.c. Just in case, I'll restate the rationale for patching it. Here is the interesting bit :

1131                 /* Ignore files which don't have the correct suffix */
1132                 if (!g_str_has_suffix (filename, G_MODULE_SUFFIX))
1133                     continue;

What it does is scan the extensions directory for *.so and then use a glib (or gobject I don't know) facility to load the extension. It makes sense for a linux program. What the patch does is override the definition of G_MODULE_SUFFIX to .dylib, so that midori finds the libraries.

What throwing bundles at this will add ?

[afb@…]

Replying to elelay@…:

Meanwhile, midori 0.2.7 has been released.

Extensions are *supposed* to have the .so suffix on Mac OS X too, but I'm OK with the proposed hack (.dylib) to get things going...

[elelay (Eric Le Lay)]

ok, ​now I get the distinction between libraries and modules. Now I don't understand how the shared libraries are loaded as extensions ??

Patching wscript to load the osx tool and extensions/wscript_build to set obj.mac_bundle = True and wscript again to set the bundle extension to .so instead of .bundle seems to work. I'll submit a patch tomorrow

[afb@…]

My MacPorts installation is currently broken, but will need to rebuild it for this commit and for the outstanding Xfce 4.6.2 upgrade also waiting to be submitted.

[elelay (Eric Le Lay)]

update midori to 0.2.7 + build bundles

[afb@…]

Fixed, r73370. (version 0.2.9)