Statistics gathering in compiler should be (can be) thread safe #9613
Labels
Milestone
Comments
|
Imported From: https://issues.scala-lang.org/browse/SI-9613?orig=1 |
|
has anything happened with this in recent work on statistics? (@jvican?) |
|
@SethTisue Nothing happened in this area because the use of statistics within the compiler didn't require thread safety. Now, with the introduction of the parallel genbcode, the use case has changed, so I'm happy to see progress on this area in the ticket that Lukas has linked to. |
|
@jvican what about gathering stats from IDE/sbt based compilations that run multiple compilations in parallel? IIRC the problem was that stats was stored in static maps that was shared across all compilations. |
|
That was fixed in 2.12.4. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In cases where compiler is used in some sort of parallel compilation (e.g. sbt) statistics (if enabled) are gathered in single place. Regardless of statistics being tempered (e.g. thread A and B wants to push/pop timer: A.push B.push, A.pop B.pop and we end in B -> A for A and A ~~> B for B ~~ instead of tow medium times we got one long and one short time) it can also fails (e.g. match error on popTimer).
The text was updated successfully, but these errors were encountered: