This avoids forcibly overriding directories that are customized by
port-specific needs.
As a concrete example, textproc/meld needs TEST_WRKSRC to be customized
so that pytest can find its tests, instead of the default meson test runner.
Details
Details
Diff Detail
Diff Detail
- Repository
- R11 FreeBSD ports repository
- Lint
Lint Skipped - Unit
Tests Skipped
Event Timeline
| Mk/Uses/meson.mk | ||
|---|---|---|
| 69 | A port may want to change its install command with meson custom build targets. I did this to keep all of them configurable if needed. | |
| 100 | I don't have a concrete use-case for this one in particular, and don't _really_ see the need to do this. However, I don't know if it'll hurt in any port's case. | |
| Mk/Uses/meson.mk | ||
|---|---|---|
| 69 | I think this should be driven by need. If you don't have a port that needs this, I'd avoid this change. | |
Comment Actions
Narrowed down review to purely need-driven changes in v2. textproc/meld needs this for its test suite to pass in a future upgrade.
Comment Actions
This is not an outright rejection, but rather would like to see the meld test suite in action first. They have still not released anything apart from the odd-numbered development release series.
Comment Actions
Sorry about the confusion. I had a patch ready to update meld locally that depended on this change. I have submitted it in the patch stack now.
Comment Actions
Hmm…not sold on this either.
meld's test suite is just pytest, which is not quite what meson's unit testing support is for. This USES is aimed for stuff that doesn't do weird things like the current major version of meld (3.22), which is effectively using setuptools with some pre- and post- dressing. The upcoming major version (3.24, in development as 3.23) has replaced setuptools with meson-python so it can build under USE_PYTHON=pep517 which specifically does not declare any test target of its own. Considering that the linked update is still on the 3.22 branch, which still runs fine, I'm not convinced to expose a conditional that would only be needed for one port and not for long.
Comment Actions
That reasoning makes sense, it's not worth it then. Abandoning this revision now. I'll change the meld update review to avoid adding a test suite.