On 2020-12-08 00:30, Magnus Ihse Bursie wrote:
On Tue, 8 Dec 2020 02:40:43 GMT, Mandy Chung <mch...@openjdk.org> wrote:
I have reviewed all lines in the patch file with or near instances of
`jdk.compiler`
Hi Magnus,
I see the motivation of moving these build files for better identification of
ownership. Placing them under
`src/$MODULE/{share,$OS}/data` is one option. Given that skara will
automatically determine appropriate mailing lists of a PR, it seems that
`make/modules/$MODULE/data` can be another alternative that skara can include
this pattern in the mailing list configuration?? I don't yet have a strong
preference while I don't consider everything under `make` must be owned by the
build team though. Do you see one option is better than the other?
@mlchung If you don't have any strong preference, I implore you to accept this
change. I **vastly** prefer to move the data out of `make`!
This is not just about Skara tooling. When working with make files, autoconf and shell scripts,
there is no fancy IDE support, so you are stuck with simple text editors and tools like `grep`.
I've lost count on how many times I've had my grep searches blow up, since I happened to find e.g.
a string in `tzdata` and get hundreds or more of hits. :-( And I do believe we will get a better
code structure if the build team "owns" `make`; or at least has a vested interest in
what's in that directory. We still suffer a lot of the old "I don't know where to put this
file, so I'll just put it in make cause nobody cares about it anyway" mentality, but I've been
working for quite some time to make that list of misplaced files shorter and shorter.
I strongly agree with Magnus for all the same reasons. To me, the data
files are clearly a form of source code that should be considered owned
by the component teams. I'm honestly perplexed over why this is being
argued against.
/Erik
-------------
PR: https://git.openjdk.java.net/jdk/pull/1611
