[
https://issues.apache.org/jira/browse/HAWQ-3?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14903248#comment-14903248
]
Michael Schubert commented on HAWQ-3:
-------------------------------------
I'd like to poke at the suggested solution a bit (assuming my understanding of
what is being proposed is correct). Why are there files that are explicitly
.gitignore'd but required for the build but not captured in the standard build
process if they are missing?
Trying to work around how an initial git clone works with respect to .gitignore
in CI (and well development) seems to solve the wrong problem. Either those
files shouldn't be .gitignore'd or if they really need to be .gitignored but
are required then make install (or some step prior) should account for dealing
with those required files no different than any other dependency may be treated.
We should not assume a developers (or CI for that matter) have a prior checked
out copy of the repository with those files but rather the build process should
assume it was git clone'd for the very first time yet still succeed without
having to rewind history prior to files being .gitignore'd.
> Missing files during initial upload
> -----------------------------------
>
> Key: HAWQ-3
> URL: https://issues.apache.org/jira/browse/HAWQ-3
> Project: Apache HAWQ
> Issue Type: Bug
> Reporter: Radar Lei
>
> As my check, there are about 60 files didn't been check in to apache hawq
> repo. That's should because they are in '.gitignore' file, so they got
> ignored while the initial upload.
> We should have apache hawq git repo get all the files in the first time, even
> they are in '.gitignore'. After that, they can be ignored.
> The most important missing ones which will cause 'make install' fails:
> tools/bin: ext
> tools/bin/pythonSrc: PyGreSQL-4.0
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
