Closed jneira closed 5 years ago
Thanks for the patch @jneira! Yes, it's a good idea to push that upstream because a race condition can happen between doesDirectoryExist
and createDirectory
if there are multiple processes/threads running it in parallel.
@rahulmutt I am afraid that now circleci test hangs without output more than 10 minutes. However maybe it is cause for another error after creating the cache directory succesfully
So I take it that this occurs only on CI (so linux)? Did you observe this locally on your Windows machine(s)? I can give this a try on my mac too.
Yeah, it doesnt hang in windows but in the file with the import hashed i always get a
Importing fieldOrderA: FAIL
Exception: user error (The import failed but it had to success. Dhall exception: C:\Users\Javier\AppData\Local\dhall\4a7866b88389e18cf481b525544fd7903325252faf3a86c8fdc981298c788a9b: openBinaryFile: resource busy (file is locked))
@jneira Can you file an issue on this with reproduction steps? It's not clear what exactly I have to run to investigate this.
Dont worry, i put it here incorrecly, it is included in https://github.com/typelead/eta/issues/915
Hi, i am getting an strange error in a circleci eta docker env:
The code being executed is:
As it only tries to create the directory when it doesnt exist i guess
Directory.doesDirectoryExist
is not working correctly. It calls in turn to Files.isDirectory which javadoc states:It behaves identically to
Directory.doesDirectoryExist
so dhall itself should fail in circleci as well: it seems in that envDirectory.doesDirectoryExist
returns false due a IOException (and not cause it really doesnt exist). No time to test it so let's see first if the patch works for etlas and then we can try to apply upstream.