Meson #170
Meson #170
Conversation
|
Switching the dist to xenial or bionic should fix the meson dependency, though the xenial one is probably too old. |
|
@jplloyd unfortunately Travis only supports up to Bionic, and Meson 0.49.0 is only provided since Disco :'( |
If meson >= 0.49 is essential, we can run the meson build on a docker image instead. Here's an example of a travis configuration using docker for the actual build: https://github.com/mypaint/mypaint-appimage/blob/master/.travis.yml |
|
I'll try that. |
|
OK, I think I had a hold of it. I'll squash the commits to have a clean history. |
|
Since the commit history will have to be cleaned up before merge anyway, I suggest you add [skip appveyor] to any further commit messages (first line), as there are currently 4 identical appveyor builds queued up. |
|
Yeah, travis succeedes, I just need to add doxygen inside the docker :) |
|
Weirdly enough, installing sphinx requires timezone configuration… So I removed the docs test. |
(should be merged separately)
|
... and when you're only changing the appveyor setup you can add [skip travis] (no underscores) to the first line of the commit message (I was a bit unclear last time - as long as it's on the first line travis/appveyor will respect it). |
|
The Gegl test on Window fails a weird way : it does not find internal Gegl libraries. Maybe that's why gegl was not installed on appveyor ? |
|
Weird, I fixed the Travis build, why did it not run again ? |
|
I'll review the changes on Saturday and then we can work on speeding up the CI. |
There was a problem hiding this comment.
Undeniably a lot more readable than the autotools setup and seems to cover everything the current setup does. Can't say that I like the level of magic in the translation build file, but I guess it does the right thing (not that I like the current level of automake magic).
The autotools setup will stick around for a while, so there's plenty of time to fix any further minor (or major) issues.
I've made a couple of comments requesting some details, but other than that;
squash the last commit, drop the shebang commit (unless there's a good reason for it) and I'll merge it.
| && ninja -C _build test | ||
| && ninja -C _build dist | ||
| && meson /sources _build_gegl --buildtype=release -Dgegl=true | ||
| && ninja -C _build_gegl test | ||
| && ninja -C _build_gegl dist |
There was a problem hiding this comment.
Assuming ninja -C _build_gegl test runs a superset of the tests run in ninja -C _build test (it should), we can skip the latter.
| @@ -1,4 +1,4 @@ | |||
| #!/usr/bin/env python | |||
| #!/usr/bin/env python3 | |||
There was a problem hiding this comment.
Why make this more constrained, when the script can run on either 2.7 or 3.5+ (and probably more)?
There was a problem hiding this comment.
IIRC python does not exist on the test machine, or Meson does not find it. I'll try to recall the precise reason.
| configuration: conf | ||
| ) | ||
|
|
||
| # TODO change generate.py |
There was a problem hiding this comment.
TODO without any detail of the change = change should be made before merge
| - sudo docker pull ubuntu:disco | ||
| - sudo docker run -t -v $(pwd):/sources ubuntu:disco /bin/bash -ec " | ||
| apt update -y | ||
| && apt install -y meson gettext libjson-c-dev libgirepository1.0-dev libgegl-dev git |
There was a problem hiding this comment.
Where in the build is git actually used, and why?
There was a problem hiding this comment.
Meson relies on Git to do ninja dist, to "recopy" the source code while preventing recopying any untracked change. It tests the commit, not the source tree.
|
Any possibility of this making any progress @jplloyd or @Salamandar ? :) |
I continued the work of @QuLogic , refactored a bit.