Add property to disable model caching #797
Conversation
oehme
force-pushed
the
oehme/disable-model-cache
branch
from
8771ae4
to
50d75b2
Compare
|
Hi @ppalaga, would that be something that you can add for us? |
|
@gnodet didn't we have a property for disabling caching for specific artifacts? |
|
@ppalaga you have a property for reloading certain class realms (e.g. for plugins that hold static state). But nothing for the model cache. |
Thanks for explaining @oehme! That's exactly, what my foggy recollection was about. |
| @@ -45,6 +45,9 @@ public SnapshotModelCacheFactory(DefaultModelCacheFactory factory) { | |||
|
|
|||
| @Override | |||
| public ModelCache createCache(RepositorySystemSession session) { | |||
| return new SnapshotModelCache(globalCache, factory.createCache(session)); | |||
| boolean enableModelCache = Boolean.parseBoolean(System.getProperty("mvnd.modelCache", "true")); | |||
There was a problem hiding this comment.
Could you please add the new property to org.mvndaemon.mvnd.common.Environment and add a bit of JavaDoc describing what it is doing?
Naming: Looking at other boolean properties, we have in org.mvndaemon.mvnd.common.Environment, perhaps the mvnd.noModelCache style would suit best here. I am not especially happy about negating names, but we already have a couple of those, and they are easier to use (given the default is with model cache enabled) from command line: -Dmvnd.noModelCache vs. -Dmvnd.modelCache=false. It is less typing. What do others think?
oehme
force-pushed
the
oehme/disable-model-cache
branch
2 times, most recently
from
46cb739
to
827ee0e
Compare
| * If <code>true</code>, the daemon will not use its in-memory metadata cache and instead re-read the | ||
| * metadata from the pom.xml files in the local repository. This is mostly useful for testing purposes. | ||
| */ | ||
| MVND_NO_MODEL_CACHE("mvnd.noModelCache", null, Boolean.FALSE, OptionType.BOOLEAN, Flags.NONE), |
There was a problem hiding this comment.
I wonder whether we should use Flags.DISCRIMINATING? Is the createCache() method called once for a given daemon instance or once per Maven build?
If it's once for daemon, I'd say it should be Flags.DISCRIMINATING. Otherwise it could happen that the user starts the daemon with mvnd -Dmvnd.noModelCache and all subsequent mvnd invocations with or without -Dmvnd.noModelCache will be served by the same daemon instance that has model cache off. That would not be correct, would it?
Flags.DISCRIMINATING will make all calls with -Dmvnd.noModelCache be served by a cacheless instance while all calls without -Dmvnd.noModelCache will be served by some other daemon instance that has the cache enabled.
There was a problem hiding this comment.
The createCache method is called every build, so it doesn't have to be discriminating. It doesn't hurt either, I don't mind :)
There was a problem hiding this comment.
If it is called every build, then no need to change anything. Thanks!
There was a problem hiding this comment.
Looking at the CI failures, we may need Flags.OPTIONAL.
oehme
force-pushed
the
oehme/disable-model-cache
branch
from
827ee0e
to
dd5a2e7
Compare
This is mostly for the integration tests of the Gradle Enterprise Maven extension, which heavily test dependency resolution and reuse the same GAVs in many test, e.g. a:b:1.0. It would be a lot of effort to rewrite all those tests and a shame to restart the daemon every time. So I wanted to ask if you'd consider this flag to disable the caching altogether.
oehme
force-pushed
the
oehme/disable-model-cache
branch
from
dd5a2e7
to
a1a9b53
Compare
Follow-up to apache#797, which didn't get backported to the mvn39 part of the code.
Follow-up to apache#797, which didn't get backported to the mvn39 part of the code.
Follow-up to #797, which didn't get backported to the mvn39 part of the code.
This is mostly for the integration tests of the Gradle Enterprise Maven extension, which heavily test dependency resolution and reuse the same GAVs in many test, e.g.
a:b:1.0. It would be a lot of effort to rewrite all those tests and a shame to restart the daemon every time. So I wanted to ask if you'd consider this flag to disable the caching altogether.