esmf-org / esmf-branch-summary Goto Github PK

After each "machine" is processed, the app switches to the "summary" branch and commits the summary.

This is by design so that files are only written in "summary" but it would be significantly faster to process each "machine" and then write to "summary" at the very end.

Switching branches can take up to a few minutes.

Specifically for repository URLs.

Freeze the first three columns: index, platform, compiler, and the first row of titles when scrolling the table. If possible, freeze the scroll bar to the bottom and right side of the table.

In the summary table, some 'artifacts' hyperlinks go to 404.

Example:
https://github.com/ryanlong1004/esmf-test-artifacts/tree/gaea/develop/gaea/gfortran/8.3.0/g/mpiuni/None

each platform branch should only contain it's own subfolder and summaries

This folder shows up in the summary branch. Though not intended, it's like due to the git add --all commands in the app.

Remove the unwanted artifacts and stop from re-occurring.

Currently the "build_passed" is the first column of the summary table.The default sorting seems to give highest priority to it, and consequently all the "0" lines come first, followed by the "1" lines. Sorting by "host" makes a lot more sense to me instead. Also, I personally expect to first see the columns with configuration information, before the columns with results, "build_passed" as the first results column.

I propose to move the "build_passed" column to sit between the "os" and "unit_pass" columns.

Further I propose renaming "build_passed" to "lib-build", and translate "0"->"fail, "1"->"pass".

Currently, the app searches for matching log and summary files dynamically by hash.

The variation below doesn't contain the hash.
https://raw.githubusercontent.com/esmf-org/esmf-test-artifacts/gaea/develop/gaea/gfortran/8.3.0/g/mpi/7.7.11/out/build.log

• Find out why the files are produced this way.
• Find out if it can be corrected from the source or needs to be accounted for downstream
• Implement

These results don't look consistent.

• 'queued' has number appended to it in some cases
• 'queued' is showing on NUPOC results others failed, likely isn't queued

The branch "jedwards_pio_update2" should be going to a new branch. It's instead going to "jedwards_pio_update".

https://github.com/esmf-org/esmf-test-summary/blob/main/jedwards_pio_update/v8.3.0b08-90-g763a128.md

The sqlite3 database is small enough I think we can try to publish it as an artifact on some kind of trigger.

Users can download and parse it however they need to.

In the summary table, add a column for the NetCDF version. This information can be determined from the build.log file. See example log file here that includes the NetCDF version.

Example output from the build log:

--------------------------------------------------------------
NetCDF library version: netCDF 4.8.0-development
NetCDF Fortran version: netCDF-Fortran 4.5.4-development

Here, the 4.8.0 (C) and 4.5.4 (Fortran) should be parsed out and put into a new column in the summary table. Review other build.log files to see if the format is consistent for other versions of NetCDF.