Is there any progress on the once-discussed idea o...
# development
h
Is there any progress on the once-discussed idea of being able to consume pre-built
native_engine.so
when working on pants itself? For those of us not working on the engine, the need to rebuilt it every time we pull is a huge productivity drag.
👀 1
👍 1
a
export MODE=debug
will save you minutes at the expense of a slightly slower pants...
☝️ 1
h
What does that do?
a
it turns off optimizations in the cargo build
i would be a ridiculously huge fan of this
would we serve it from some s3 which gets populated by a travis run?
h
So we'd still be building the engine, it would just be a bit faster?
Yeah, we only need to populate it from master
come to think of it we already create it on S3 for every build
a
that's what i was thinking
i need to go to an appointment for an hour but i'll make an issue/pr about this if nobody else gets to it first
a
On my laptop, “a bit faster” is like a 9 minute difference on a clean build…
a
is there a reason we can't default
MODE=debug
?
a
The problem with defaulting to
MODE=debug
is is slows down pants sometimes like 10x, so people trying to use pants in source mode in large repos, or do any kind of benchmarking, will get really slow results, and it may not be obvious why. People should only need to re-build the rust stuff proportional to when we make rust changes, which is still pretty rare (the big kicker is every 6 weeks they’ll need to clean build because we update rust versions)….