Sort config types by their load dependencies once#10148
Conversation
icinga-probot
bot
added
the
core/quality
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
8a2ee5a to
f78108d
Compare
julianbrost
left a comment
julianbrost
left a comment
There was a problem hiding this comment.
Compared to #10003, this PR has no cycle detection (at least I believe the purpose of some of the code might be cycle detection, it's not entirely clear though: #10003 (comment)). When I intentionally introduced a dependency in the *.ti files, nothing in mkclass stopped me from doing so and all I got was a not very helpful error message in a test (see inline comment). So some more explicit "this failed due to a cycle" error reporting/VERIFY() might be a good idea.
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
f78108d to
7c56ec3
Compare
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
3 times, most recently
from
bfcfe80 to
e019556
Compare
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
e019556 to
e011597
Compare
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
2 times, most recently
from
b21a41f to
2c308dd
Compare
yhabteab
requested review from
julianbrost
and removed request for
julianbrost
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
2c308dd to
9442803
Compare
julianbrost
left a comment
julianbrost
left a comment
There was a problem hiding this comment.
Looks fine and should do the job, I just don't want to do a quick approve + merge right before leaving for vacation and another pair of eyes shouldn't hurt. But I'd say you can optimistically adapt #10000 already.
I'm not 100% happy regarding the needs for the test changes including the CMakeLists changes, but that's probably what happens when relying on linker/loader magic for the type registration (the alternative would be to perform an extra self check within the implementation, so that it's checked every single time, but that's also not too nice).
julianbrost
dismissed
their stale review
No more open requests for changes, see previous comment.
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
9442803 to
abb068e
Compare
|
Oh! We can't backport this to 2.13.x due to 33e609d. |
|
Do you mean "We can't backport this to 2.13.x without 33e609d"? |
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
abb068e to
b85fc9f
Compare
|
Just rebased it! |
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
b85fc9f to
cb4fe57
Compare
…es once to avoid complicated nested loops, iterating over the same types and checking dependencies over and over, skipping already completed ones.
yhabteab
force-pushed
the
enhanced-sort-types-by-load-dependencies
branch
from
cb4fe57 to
eb97676
Compare
| types/instantiate | ||
| types/sort_by_load_after | ||
| ) | ||
|
|
Since you're back now :) and the PR is still open, do you have any change requests? Otherwise, please give your approval as well. |
3 participants
This PR is basically the same as #10003, but since @Al2Klimov is out for this week, we need to get it done. So I've cherry-picked the commits from there and added some additional commits to address the requested changes in that PR.
closes #10003