Originally part of #17, but #17 has a bad title and half of the issue is fixed already. Hence this new issue:
To save time during validation, we currently only check selected DC files (depending on docserv-dchash's deduplication of DC files). For the most part, this is great, however, we won't notice when a DC file that is not checked references a ROOTID that does not exist at all or does not exist in the current profiling.
To solve, this, I think we should:
Make a secondary list of dependent documents to validate, if there are any. The first document listed on each line would be the one on the actual build list, every other DC file listed on that line are the ones that were deduplicated away
Validate the list of secondary documents in the same runner that we validate the DC-*-all in, taking advantage of existing profiling etc. (We could also go via a bigfile and grep through it, but that probably would not yield gains while increasing the amount of code we need to maintain for our action.)
Originally part of #17, but #17 has a bad title and half of the issue is fixed already. Hence this new issue:
To save time during validation, we currently only check selected DC files (depending on
docserv-dchash
's deduplication of DC files). For the most part, this is great, however, we won't notice when a DC file that is not checked references a ROOTID that does not exist at all or does not exist in the current profiling.To solve, this, I think we should:
Make a secondary list of dependent documents to validate, if there are any. The first document listed on each line would be the one on the actual build list, every other DC file listed on that line are the ones that were deduplicated away
Validate the list of secondary documents in the same runner that we validate the DC-*-all in, taking advantage of existing profiling etc. (We could also go via a bigfile and grep through it, but that probably would not yield gains while increasing the amount of code we need to maintain for our action.)