[r/ci] Point at correct TileDB-R location #2275
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #2275 +/- ##
==========================================
- Coverage 80.28% 78.89% -1.40%
==========================================
Files 87 139 +52
Lines 6367 10718 +4351
Branches 214 215 +1
==========================================
+ Hits 5112 8456 +3344
- Misses 1169 2164 +995
- Partials 86 98 +12
Flags with carried forward coverage won't be shown. Click here to find out more.
|
|
[sc-43172] |
|
This pull request has been linked to Shortcut Story #43172: Update projects to TileDB 2.21. |
.github/workflows/r-ci.yml
Outdated
| # sources --- but that would require installing all build dependencies. To see what the most | ||
| # recent binary is run e.g. docker run --rm -ti rocker/r2u bash -c 'apt update -qq && | ||
| # apt-cache show r-cran-tiledb' Using the r-universe builds (as below) is a suitable fallback | ||
| # as they update more frequently than CRAN. |
There was a problem hiding this comment.
I am not really sure I agree with the wording. Sadly I cannot quite put my finger on something better.
But riffing for a moment we seem to have the three situations here:
- tiledb-r and tiledbsoma agree on versions, the world is good
- tiledb-r is
- at release at CRAN (which is what we generally want)
- at release-candidate in the GH repo and available via selected r-universe builds
That last bit is "unusual but common" and I think the bit we have here today.
And just how he have / had the existing ("sleeping") PR #1994 we have this now. Part of me thinks we may want to discuss how to drive this better from one top-level (possibly outside the repo?) 'declarative' file.
It's tricky.
There was a problem hiding this comment.
Can you remove the new leading paragraph? I do not think it illuminates sufficiently.
What we have here is that sometimes we need jump a step ahead and sometimes we need to not do that. So we comment these stanza 'in' and 'out' as they give us the binaries we otherwise want to speed things up.
@eddelbuettel I put it in a while back because I found it to be necessary information for me. And I do release management for this project -- so I believed that was important. Then, on another commit (not by me), that paragraph disappeared. Then, on this commit I'm trying to put it back in. Here, you're asking me to take it back out again. What I will do is to take the wording I find crucial -- to avoid CI breakages like the one we're having right now -- and I will put that wording at https://github.com/single-cell-data/TileDB-SOMA/wiki/Branches-and-releases, and I will add that hyperlink as a comment within the |
|
@johnkerl My comments in this ticket were aimed at illustrating the complexity of the issue. The comment in the yaml did not reflect it to my reading, but focussed on a different (and here less relevant) issue. No more no less. It is by our chosing that we sometimes want what r-universe offers, and sometimes we don't. I maintain that a more central switch may help us all do our job and more easily. The key aspect remain that we sometime do, and sometimes do not, need these two chunks. Right now all we need was to actually comment the other way to get to prior release of tiledb-r with the (pinned as fallback) tiledb core version used here. That toggle will always have be moved forward or backwards depending on where the SOMA release is relative to the tiledb-r release. But that has actually little to nothing to with where we get binaries from to speed up the CI. It really is just a complicated form of pinning of tiledb core. As such I found the comment misleading. |
* [r/ci] Point at correct TileDB-R location * Restore valuable comment [skip-ci] * code-review feedback
* [r/ci] Point at correct TileDB-R location * Restore valuable comment [skip-ci] * code-review feedback Co-authored-by: John Kerl <[email protected]>
See comments in the affected CI YAML file.