Re: Cannot build 3.16.2 on Windows

[Kai Willadsen <kai willadsen gmail com>]

On 14 June 2017 at 21:37, Serhan Apaydın <zerhan gmail com> wrote:
> Hi,
>
> I'm trying to build Meld on Windows. I've cloned the repo, installed Python
> 2.7.6, cx_Freeze 4.3.2 and PyGI 3.14.0 AIO. The description on
> https://download.gnome.org/binaries/win32/meld/3.16/SOURCES states that we
> should have PyGI 3.14.0 AIO rev19 installed but I couldn't find rev19
> anywhere. There's only rev22 (pygi-aio-3.14.0_rev22-setup.exe) on
> https://sourceforge.net/projects/pygobjectwin32/files
>
> If I install rev22 and execute `python setup_win32.py bdist_msi` I get some
> "libgnutls-26.dll is missing" error.

I'm quite confident that I didn't need this step. It's possible
however that selected something more than the default in the AIO
installer... unfortunately the build machine that I was using is no
more, so I can't check.

> I've manually copied that file into "Python27/Lib/site-packages/gnome"

Where did you get this library from?

> and executed `python setup_win32.py
> bdist_msi` again, sucessfully created the msi file, and then I installed
> Meld using that msi but Meld failed to launch. It gives the following error:

Just so you know, you should also be able to run Meld uninstalled at
this point. If that works but the MSI doesn't, then it's basically
going to be a packaging issue. If it doesn't work, then chances are
good that it will be easier to debug running uninstalled.

<snip>
>   File "bin/meld", line 259, in <module>
>   File "bin/meld", line 162, in check_requirements
<snip>

So this line is our gobject introspection GTK+ import, so something in
that chain is failing. It's annoying that we don't get more details
from cx_freeze/python here, but hey. My best guess right now is that
the libgnutls DLL is somehow mismatched with the rest of the DLLs,
but... honestly I'm really not sure.

> Is there anything I should do? Is it because I've installed rev22 instead of
> rev19? If so, where can I find rev19? Just in case I've created a ticket and
> asked the author to re-upload rev19:
> https://sourceforge.net/p/pygobjectwin32/tickets/55/

While I can't rule out a problem here, the AIO binaries have been
fairly reliable, so I'd be a bit surprised if this was the cause.

cheers,
Kai