We were already avoiding re-processing a same DLL within the same run
(this can happen when 2 dependencies have themselves a common
dependency). But the dll_link.py script was stateless regarding previous
runs so we might be checking again the same DLLs multiple times (even
though we were not copying them again).
Let's make the script stateful with a new parameter to give a file where
all the previously processed DLL names are stored. I am hoping it would
improve the efficiency of the packaging-win32-native which is suddenly
extra slow (it always times out, even after raising the max job time;
now we time out after 2h30! The 64-bit packaging job just takes 1h,
which is too much already, but still much more reasonable).
Also use it to fix packaging of GIMP for the Windows installer (native
CI job). The CI was indeed failing to package libbrotlienc.dll,
dependency of libjxl.dll, for the simple reason that they were on
different prefixes. By calling dll_link.py on one prefix, then the
other, we were failing to grab the deeper dependency. Now with this new
ability to set several sources, the script is able to search everywhere
(with first prefix given on the CLI call as priority).
The MSYS2 package got recently bumped from 3.8 to 3.9.6.
At first I wanted to update our packaging and installer scripts to be
more generic using glob patterns (so that they should work now and
should continue to work even if bumping to a higher minor version in the
future). Unfortunately this would work for `package-gimp-msys2.sh` but
in `files.isi`, it would only work for `libpython3.*.dll`, not for the
python3.9/ folder. InnoSetup apparently doesn't support using a folder
as source (or maybe just a folder with glob like `python3.*`) as it
resulted in a "No files found matching" error.
So leave everything with the accurate version (because anyway it's much
better to get an early failure than only at the very last step).
Since gio searches its modules based on the paths as advertized by the
pkg-config, let's just move the pre-compiled modules (by MSYS2
packages). We build the same version of glib2 with the same options, and
only one additional patch. So this should not be a problem to use the
pre-built modules rather than rebuilding glib-networking too.
Since we moved most of them to bin/, share/lua*/ and lib/lua/ files are
not necessary anymore (according to my tests so far at least). Let's not
include them.
Also exclude the lua DLL from the generic libraries. It is only for when
the lua option is set.
It looks like the DLL ends up searched into bin/lgi/ and other lua files
from bin/lua/. There might be better solutions for the issue, but for
the time being, it seems to work ok.
Note that ender used instead to rebuild lua with the following changes
(cf. IRC):
> src/luaconf.h:
> #define LUA_PATH_DEFAULT \
> "!\\..\\share\\lua\\5.1\\?.lua;" "!\\..\\share\\lua\\5.1\\?\\init.lua;" LUA_LDIR"?.lua;" LUA_LDIR"?\\init.lua;"
> #define LUA_CPATH_DEFAULT \
> LUA_CDIR"..\\lib\\lua\\5.1\\?.dll;" LUA_CDIR"?.dll;" LUA_CDIR"loadall.dll"
But moving files around in the installed tree is much simpler than
rebuilding the whole lua just for this.
Note that I must not install both lua51 with luajit because these are
conflicting. Let's go with luajit from feedback we had on the best
choice (though this topic itself seems a bit heated actually).
Also clean-out the unexpected file removal because now I had the
opposite case, i.e. a CI problem because of this. And from my latest
tests, it seems to work ok for the time being without, which is much
cleaner anyway. So let's go like this for the time being.
Add GObject Introspection libs, python libs and various binaries. Not
sure why we need 3 python binaries. This would need to be investigated
and possibly cleaned up a bit. Similarly maybe it would be better to be
a bit more selective about what we take from lib/python3.8/ folder
(looking what's in there, I wondered if some of the stuff were not
pulled by some dependencies and/or should maybe be filtered out).
But for now, let's make the InnoSetup scripts happy.
In any case, the resulting installer was tested and now Python plug-ins
work successfully. Wouhou!
Run InnoSetup in the Windows CI to build the installer from both the 32
and 64-bit builds.
Current limitations:
- No installer signature yet.
- Dependencies will have to be checked more thoroughly.
- Apart from babl and GEGL, we may want to make custom builds of any
package which has a patch in build/windows/patches/ (Windows-specific
patches) and build/patches/ (all platform patches).
- Plug-in interpreters (Python, Lua…) don't work. This will need to be
looked at in detail.
Globally this first automated installer build works fine though, as I
could install it in a Windows 10 VM and GIMP ran fine! So it's a first
step towards fully automated releases for Windows.
These files I'm adding are currently being requested by the installer
script, even though I'm not sure at all how much needed they are,
especially for some of them (because I could successfully create an
installer by just commenting these out in the installer script).
This new job resulted in a package which allows to run GIMP on Windows
(as tested in a VM; at least it starts, I can create a new canvas and
paint). Of course I think this will need to be tweaked a little bit
more, as I'm sure we miss things here and there.
At the very least, even though I add the Python and Luajit binaries,
GIMP on Windows didn't find them. This will need to be investigated.
Also it looks like opening from a remote location may not work. Not sure
if this about a missing GIO module or maybe something which works
differently on Windows (I was not even able to drag'n drop from the
browser!). Anyway this needs to be looked at as well.
Note that gdk-pixbuf-query-loaders is apparently unneeded when GIMP is
built this way (unlike with our crossroad build).
All this to say that this is still an early attempt to full CI build for
Windows.
It doesn't invalidate the crossroad build, because cross-compilation
builds from Linux will always stay very important for Linux developers
to be able to easily fix Windows bugs too; yet the crossroad build has 2
major issues:
1. We haven't figured out yet how to run GObject Introspection tools for
cross-builds, so the crossroad builds are not full-featured (and this
is quite a major feature we are missing!).
2. Also I will want to run the installer in the CI at some point and the
one we use can only run on Windows itself AFAIK. We could try to run
it through Wine, but still anyway the point 1. is already quite a
blocker, let's do the simple thing.
Note that we will likely want to move to meson for this build, because
autotools is very slow on Windows. But as long as the few blocker meson
bugs are not fixed, let's stick to the slow yet good build.
