GIthub actions docker containers (for windows build)

[Dirk Hohndel]

I'd start with b, then d

5.15 is something we'll need to get to eventually.
In the past there was an MXE contributor who tended to work on that, maybe reach out to them as well.


/D


On May 28, 2020 9:39:32 AM PDT, Paul Buxton <paulbuxton.mail at googlemail.com> wrote:
>So down the rabbit hole!
>
>Taking a reasonably up to date version of MXE we hit a bug in Cmake
>3.15
>where when specifying a PKG_CONFIG_PATH with multiple folders in the
>path
>and cross compiling to windows (on a Linux host) it incorrectly swaps
>the :
>with a ; so pkg-config only searches the first entry on the path.
>This issue has been fixed in CMake 3.17.
>
>So should I,
>a) Try getting cmake manually of a version we choose
>b) Patch the cmake we have in the container (it's a 1 line fix)
>c) Get MXE to use the newer cmake
>d) Get MXE to cherry pick the fix
>e) Something else
>
>Options (c) and (d) would also end up pulling in QT 5.15, and when I
>briefly tried this yesterday it was failing to build qtbase...
>Option(a) might have other unknowns.
>Option (b) is a bit hacky :-)
>
>Any preferences?
>
>There is also an issue with missing symbol
>libglib is missing _imp__timeGetTime
>Which seems to be related to this change (or something similar) in the
>Mingw toolchain
>https://sourceforge.net/p/mingw-w64/mingw-w64/ci/65042f860f8cb67b26d97fd763143946f9a7c6e6/
>
>But I haven't figured out yet where mxe gets its mingw from, so haven't
>been able to oinpoint when/how this has been broken.
>
>
>
>
>
>On Thu, May 28, 2020 at 6:59 AM Paul Buxton
><paulbuxton.mail at googlemail.com>
>wrote:
>
>> Yeah, I saw the parameter and an using it locally, but as Dirk
>mentions it
>> isn't used by the build on GitHub.
>> I will include this in my changes when I get something that behaves.
>>
>>
>> On Wed, 27 May 2020, 20:00 Dirk Hohndel, <dirk at hohndel.org> wrote:
>>
>>> Hi Salva,
>>>
>>> you are of course correct. And I had used a script to pass that in
>here,
>>> but then never checked in that script.
>>> As a result the GitHub Action happened to use 'master' - which
>wasn't
>>> really what I had intended.
>>>
>>> /D
>>>
>>> On May 27, 2020, at 10:52 AM, Salvador Cuñat
><salvador.cunat at gmail.com>
>>> wrote:
>>>
>>> Hi Dirk, Paul
>>>
>>> El mar., 26 may. 2020 17:30, Dirk Hohndel via subsurface <
>>> subsurface at subsurface-divelog.org> escribió:
>>>
>>>>
>>>> The MXE version is fixed as a specific SHA. And as I grep through
>the
>>>> sources to refresh my memory where that's done I notice that the
>SHA
>>>> doesn't appear to be in the source tree. How weird.
>>>> I believe the last one that I used (based on the remnants of local
>>>> builds) was 9f6b9c6f - but I need to dig some more to figure out
>why this
>>>> isn't in the sources...
>>>>
>>>
>>> In aab658f we introduced a parameter for the docker build to pass
>the mxe
>>> git sha using  --build-arg=
>>>
>>> After this change the used version is not in the Dockerfile any
>more, but
>>> it's manually entered while building the image.
>>>
>>> Best regards.
>>>
>>> Salva.
>>>
>>>
>>>

-- 
From my phone
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.subsurface-divelog.org/pipermail/subsurface/attachments/20200528/051529c3/attachment-0001.html>