possibly it shouldn't do that for a non-trivial comprehension though
and disitributed bulds that ca leverage a cluster
I am evaluating different build system needs to support distributed builds, cross-compile toolchain and caching. Bazel is one of the canndaidates and but not fond of having to have jre and its startup and memory implications
I like please seems similar but written in golang. But I can't tell much about how actively the developmennt community is around it
The functionality we need, that I can't seem figure from the documentation , if it is supported ie. please can build. 1. cross compiling different CPU architectures. 2. Building a platform/package that can contain multiple libraries/components meant for different CPU architectures. 3. Distributed Builds that can be run on multiple remote compute nodes(This, I think is "remote" and is marked experimental
it's stopped being a priority for us internally but support isn't going to go away (although it might need some more polish in places)
Distributed builds are in active development at the moment
unfortunately it's taking a while, among other things we decided to use Google's RE API for it which has been very complicated to integrated
Take a look at the pleasing js repo, https://github.com/thought-machine/pleasings/tree/master/js . You can use them with
subinclude('///pleasings//js'), see https://please.build/pleasings.html
@peterebden could you push a new release for please, we could benefit from the #870 to make coverage useful with tools like codecov
thanks
have been tracking a bug in the last release, but the fix is taking too long so I'll roll that back for now
afaik it caches results based on stuff in
data and the test script generated by cmd
the
data doesn't seem to respect need_transitive_deps
I was thinking I could use
$(hash ...) to force my .sh file to change, but it doesn't account for transitive deps too
it's easy to do for regular builds,
needs_transitive_deps work well but when I start dealing with binaries that are run during build it gets messy
it would be great if someone could point me in the right direction, especially with the
gentest thing
It doesn't work :/ It reruns the
cmd, but it produces the same .sh no matter how transitive deps look like. So when it runs the test, the cmd output is the same as before, the data is the same as before (it doesn't count transitive deps) and it reuses the cached result instead of running the test again
I think a method similar to
$(hash ...) would be very useful, if it would return a hash of the input, not of the output
Then I could add it to the
cmd to include the hash in the .sh
I suppose we could add a
needs_transitive_data flag as well
you might be able to get the effect you want using
stamp = True, needs_transitive_deps = True and echoing $STAMP into the binary
I think that would work
if you made say a python_binary stamped it wouldn't do anything with the variables