reduce file access #4461
reduce file access #4461
Conversation
At least on Windows the repeated invocation of file.exists takes a huge amount of time.
| @@ -68,10 +68,13 @@ public File getLocation(boolean fetch) { | |||
| if (projectLocation != null) { | |||
| return projectLocation; | |||
| } | |||
| if (fetch && locationSupplier != null && (location == null || !location.exists())) { | |||
| if (fetch && locationSupplier != null && location == null) { | |||
There was a problem hiding this comment.
The check was there to make sure the file was not deleted in the meantime (e.g. outside of control of the descriptor), this won't be detected anymore.
There was a problem hiding this comment.
I'm aware of that. But I believe the improved performance is totally worth that. Also, if this is related to the "file is corrupt, trigger download again" functionality, then I think we should do something there were the download is re-triggered. here it really costs far too much. In the above screenshot it's 6 minutes in this single method call.
There was a problem hiding this comment.
Also, if this is related to the "file is corrupt, trigger download again" functionality, then I think we should do something there were the download is re-triggered
The problem is that it is not that easy. We know that the file is corrupt (and can delete it) but there is no easy way to know here it was downloaded from.
Test Results 603 files 603 suites 4h 4m 21s ⏱️ Results for commit 880caa4. |
At least on Windows the repeated invocation of file.exists takes a huge amount of time. The change removes several minutes of runtime from a 1 hour product build:
