Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

OSGeo4W workflow for GitHub Actions #692

Merged
merged 64 commits into from
Jul 2, 2020

Conversation

HuidaeCho
Copy link
Member

@HuidaeCho HuidaeCho commented Jun 3, 2020

Add a new workflow for compiling GRASS GIS with OSGeo4W and MSYS2 on Windows. It also runs unit tests using the grass.gunittest.main module and nc_spm_full_v2alpha2 data set.

@HuidaeCho HuidaeCho added help wanted Extra attention is needed windows Microsoft Windows specific labels Jun 3, 2020
@HuidaeCho HuidaeCho marked this pull request as draft June 3, 2020 15:28
@neteler neteler changed the title [WIP] OSGeo4W workflow [WIP] OSGeo4W workflow for GitHub Actions Jun 3, 2020
@wenzeslaus
Copy link
Member

@HuidaeCho The workflow in this PR runs and "succeeds", so I don't even see any errors. The OSGeo4W workflow in Checks says: OSGeo4W / windows-latest build succeeded 5 hours ago in 37s. I don't know how osgeo4w-setup-x86_64.exe and PowerShell behave. Is a non-zero exit code being ignored here? Given that almost all workflows in your repo were canceled for the latest commit on your fork, I would say there might have been some glitch. You can always re-run if you are in doubt and see then.

@HuidaeCho
Copy link
Member Author

@HuidaeCho The workflow in this PR runs and "succeeds", so I don't even see any errors. The OSGeo4W workflow in Checks says: OSGeo4W / windows-latest build succeeded 5 hours ago in 37s. I don't know how osgeo4w-setup-x86_64.exe and PowerShell behave. Is a non-zero exit code being ignored here? Given that almost all workflows in your repo were canceled for the latest commit on your fork, I would say there might have been some glitch. You can always re-run if you are in doubt and see then.

That's interesting. It actually worked on this repo! I think this issue on my repo is somehow related to all those error messages I'm receiving whenever I merge upstream/master into my master. You don't have this problem on your repo? Can it be my setting? Or maybe, personal repos are different from organizational repos? Confusing. I didn't want to test my github skills in this PR, but looks like I don't have any other options now. :(

@wenzeslaus
Copy link
Member

It actually worked on this repo! I think this issue on my repo

If it works here, you can just continue adding things and see how that develops. Did you try to re-run the jobs?

As for the success here. Is the return code actually checked? Do you see expected results?

is somehow related to all those error messages I'm receiving whenever I merge upstream/master into my master.

I don't think it should be related to how you are using Git. What error messages? The commits here and on the branch look good (both are and should be the same anyway). However, you should not be merging into your master if you do, use rebasing instead. See CONTRIBUTING.md.

You don't have this problem on your repo?

No. However, there are some problems that in forks we get that tests on Ubuntu 16.04 just take too long (>6 hours) and are killed. Never happened on master branch yet.

Can it be my setting?

I'm using the defaults, but there is not much to set anyway. You can disable yours and just see them in the PR if that really bothers you, but I would first see if there is a pattern.

Or maybe, personal repos are different from organizational repos?

If something, it would be more forks versus original repos.

@HuidaeCho
Copy link
Member Author

HuidaeCho commented Jun 6, 2020

gunittest fixes

@HuidaeCho
Copy link
Member Author

HuidaeCho commented Jun 7, 2020

test_thorough.bat times out (6 hours limit) at test_r_blend.py? Anyway, the workflow itself seems to work fine now.

@wenzeslaus I tested it on my Windows machine. test_r_blend.py opens a new monitor and it seems to wait for the monitor to be closed (?). The moment I closed the monitor, the test moved to the next one. Changed wx0 to the non-interactive png monitor in 517c808. It will create testreport/scripts/r.blend/test_r_blend/map.png.

@HuidaeCho HuidaeCho added enhancement New feature or request and removed help wanted Extra attention is needed labels Jun 8, 2020
Copy link
Member

@wenzeslaus wenzeslaus left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is awesome!

The test in r.blend is not well written to start with, but you fix makes it not harmful, so I second that. The encoding seems seems to be working and it makes sense to me, so I would go with that. The YAML file and the scripts seem to be split reasonably. I'm adding some minor comments here and there.

One thing to double check is that it really fails if the tests fail. I think you confirmed it fails at one point, but I said, double checking, because I don't think we will be examining the textual output for each PR/commit especially once we increase our success rate there in the future.

However, I myself don't compile on Windows, so it would be good if someone else who compiles GRASS GIS on Windows adds a second review. (@landam @hellik ?)

lib/python/gunittest/reporters.py Outdated Show resolved Hide resolved
.github/workflows/osgeo4w.yml Outdated Show resolved Hide resolved
.github/workflows/test_simple.bat Outdated Show resolved Hide resolved
.github/workflows/test_thorough.bat Show resolved Hide resolved
@HuidaeCho
Copy link
Member Author

One thing to double check is that it really fails if the tests fail. I think you confirmed it fails at one point, but I said, double checking, because I don't think we will be examining the textual output for each PR/commit especially once we increase our success rate there in the future.

Assuming that each and every unit test script is written well and reports correct results to grass79.py, I confirmed that the workflow actually fails when the success rate drops below --min-success.

I'm just wondering why around 20% tests are still failing on Linux builds. 40% failure on OSGeo4W is really poor, but it's understandable (?)... Are we 100% confident that all unit tests are written well?

@HuidaeCho HuidaeCho requested a review from wenzeslaus June 9, 2020 13:27
@wenzeslaus
Copy link
Member

Assuming that each and every unit test script is written well and reports correct results to grass79.py, I confirmed that the workflow actually fails when the success rate drops below --min-success.

Yes, grass...--exec passes the non-zero return code of whatever is being executed and gunittest now newly return non-zero when at least one test fails. The tests failing or not that's of course another story, but here we can just assume it works.

I'm just wondering why around 20% tests are still failing on Linux builds. 40% failure on OSGeo4W is really poor, but it's understandable (?)...

Ideally, we would have 100% tests passing and it is both realistic and reasonable requirement. The tests running on my Linux server were above 90% if everything was all right (2014-08-01 till 2019-03-26):

http://fatra.cnr.ncsu.edu/grassgistests/summary_report/nc/index.html

Unlike the Python code checks, I did not review the tests in detail after adding them to GitHub Actions, so there might be some trivial problems. From running them on my machine, I can see from the reports that there is a group of tests failing in relation to band references in temporal framework. The errors don't show in the GitHub Actions, but that's fixable.

Are we 100% confident that all unit tests are written well?

As you saw with the r.blend test, no, but that one is also extreme case which slipped review during Google Code-in. Some tests are from the pre-test-automation era, so they may or may not work and are waiting to be corrected. Other than that, it is really case by case and examining the failing test is necessary.

@wenzeslaus
Copy link
Member

@hellik @landam It would be great to have at least one additional review for this (see the note in my review).

# and create bin.x86_64-w64-mingw32\grass$ver.bat (run this batch file to start
# GRASS GIS) and dist.x86_64-w64-mingw32\etc\env.bat.
#
# -p optionally install GRASS GIS to C:\OSGeo4W64\opt\grass (run
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are the binaries downloadable somewhere? could these be used to upload for daily builds in OSGeo4W?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not at this point, but we could use GitHub artifacts to copy out the package: https://help.github.com/en/actions/configuring-and-managing-workflows/persisting-workflow-data-using-artifacts

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok.

if we're going this way later on, maybe the install scheme (now in \opt) have to be adapted to be usable in OSGeo4W?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The current workflow is mainly for testing, but changing the install scheme wouldn't be a problem. I thought about that possibility too, but we already have working daily builds for Windows. Also, I wanted to create a portable package that doesn't require admin rights to install.

@hellik
Copy link
Member

hellik commented Jun 23, 2020

@wenzeslaus :had a quick review regarding windows compilation, looks good so far. if we're using it later on with GH artifacts (and OSGeo4W daily builds), maybe some fine tuning is needed.

@wenzeslaus wenzeslaus removed the request for review from landam July 1, 2020 00:46
@wenzeslaus
Copy link
Member

Thanks @hellik for review. We won't be waiting for @landam and merge.

@HuidaeCho can you please change the PR description to reflect what have you done so we can use it as a commit message?

@wenzeslaus wenzeslaus added the CI Continuous integration label Jul 1, 2020
@HuidaeCho HuidaeCho merged commit 7a5e933 into OSGeo:master Jul 2, 2020
@HuidaeCho HuidaeCho deleted the osgeo4w_workflow branch July 12, 2020 03:33
@HuidaeCho
Copy link
Member Author

PR for artifact archiving in #1212.

@neteler neteler added this to the 8.0.0 milestone Dec 9, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CI Continuous integration enhancement New feature or request windows Microsoft Windows specific
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants