softwarepub
Issues found during manual testing of 0.8.0 at deRSE24
- [x] No log output?
- [x] No complaints about invalid CFF file -> Logs are not enabled by default, missing init to ERR/WARN/INFO.
- [x] No complaints about non-existent or missing configuration options (example: harvest.cff.validate -> harvest.cff.enable_validation & harvest.from -> harvest.sources renamed) (Potential duplication of above)
- [x] #259
- [x] #260
Version given as 1.0 in CFF is interpreted as floating point (YAML parser!) and will give an error at deposit because not a valid string. We need to do a hard cast to string for the version
Logs and errors are now handled a bit saner... i.e., the errors are output to the log.error channel (which is by default printed on the console).
Still lot's of stuff to do that will be handled in different issues (i.e., #238).
During the workshop, it was also mentioned that hermes should support older Python versions (i.e., 3.8).
I'm strictly against this, seeing that EOLs are approaching fast. 3.8 will be unsupported from October 2024, 3.9 October 2025.
I'd argue that requiring 3.10 is actually feasible and shouldn't be changed.
I.e., whoever is working on the 3.8 backport (mentioned in the paper draft) should drop their pen IMHO.
I think 3.8 was also required to support SOMESY
True. I suggest asking somesy to update to 3.10 instead of supporting versions near the EOL. Or were there any (dependency-related) reasons for sticking with 3.8 (other than: a lot of people still use it)?
I think we can close this issue. Python 3.8 is EOL now and even 3.10 is going to be EOL in 1 year. See also https://endoflife.date/python
The logging stuff is tracked via #138, too.
The casting stuff is tracked via #281 / #291 .