Fix Version/s: None
Component/s: buildbot, lsstDoxygen
The buildbot scripts have an explicit dependency on the datarel package, which we'd like to remove from the stack. It uses datarel as the top-level product when building the cross-linked HTML documentation; lsstDoxygen's makeDocs script takes a single package, and generates the list of packages to include in the Doxygen build by finding all dependencies of that package.
So, to remove the explicit dependency on datarel, we need to either:
- find a new top-level product with a Doxygen build to pass to makeDocs (e.g. by adding a trivial Doxygen build to lsst_distrib)
- modify the argument parsing in lsstDoxygen to take a list of multiple products (it looks like the limitation to one package is only in the argument parsing), and pass it a list of top-level products in the buildbot scripts.
This is currently a blocker for
DM-2928, which itself a blocker for DM-1766, which has now been lingering for a few weeks now. I'm going to look for other ways to remove the block on the latter, but I don't have a solution yet.
DM-2928 move old ingest scripts into and retire old packages
- Won't Fix
- is triggered by
RFC-57 retire datarel, ap, and testing_endToEnd packages
- relates to
DM-6199 Stack API documentation
DM-9458 Port makeDocs to python 3
DM-13074 doxygen.lsst.codes master index.html page broken
DM-7623 Modernize tests in datarel to support pytest
DM-8260 Remove scisql dependency from datarel
DM-7298 Port datarel to Python 3
- Won't Fix
RFC-449 Add GnuTLS as a macOS platform prerequisite
I'm going to do this in the near future, yell if you have any concerns not captured in the ticket.
Is this going to fix itself when Jonathan Sick switches us away from makeDocs?
I've just linked this to the Pipelines doc epic so I can remember to close out this ticket when docs are being published that way. I don't have more specific tickets yet.
I don't think we can really mark this as invalid until we fix makeDocs or migrate the document building to the new system.
I have done some quick tests with makeDocs and it seems to me that we can switch from datarel to lsst_distrib without any need to change lsst_distrib. In my test I get a full set of documents if I ask for lsst_distrib (including ctrl_pool which is missing from datarel.
I'm still setting up a stack for this, so you're one step ahead of me Tim Jenness.
If it's a python 3 stack you'll need
I have a branch that removes datarel from the doc build but I don't think it has ever been tested. I'll rebase it and open a PR.
The branch replaced datarel with lsst_distrib https://github.com/lsst-sqre/ci-scripts/pull/44 but someone needs to test that makeDocs works with lsst_distrib (or any other product) before that's merged.
One problem is that
DM-13074 is indicating that we may see breakage regardless.
Are you OK with disabling the doc build if that is the case? Otherwise, it will break the nightly/weekly.
I'm not sure what you are asking. Doc builds are broken (
DM-13074). In my tests a year ago everything worked fine when I switched datarel to lsst_distrib but I can't imagine changing this will fix DM-13074 (and I think on that ticket you said you had made the switch as part of debugging). Switching to lsst_distrib shouldn't make things worse though.
Currently, the doxygen landing page is incorrect but the doc build is still running. I believe all the classes, etc. are still there if navigated to by the index. If the doc build starts failing, as in exiting non-zero, it will have to be disabled completely.
Given my test from a year ago I believe switching to lsst_distrib will not be a problem.
OK - I have merged that PR. I'm currently re-run last nights release, so it will probably be several hours before this can be tested via release/run-rebuild.
Joshua Hoblitt I have removed datarel from lsst_distrib and your change to makedocs has worked. Feel free to close this ticket (that was opened nearly 3 years ago)
Tim Jenness I think
DM-7298 can be invalidated now.
Lowering the priority as I think I've found another way to unblock