scalajsc fails on windows upon first invocation

[gzm0]

The first invocation after scalajsc has been added to the path, it seems to not find the Scala.js library (log follows). The next invocation works as expected.

How to reproduce (use foo.scala from test cases):

> set PATH=%PATH%;<...>
> scalajsc foo.scala

[gzm0]

Test log

>set PATH=%PATH%;...\scalajs_2.11\bin;...\scala-2.11.1\bin

>cd tmp

tmp>dir
 Volume in drive E is PersistentDataDisk
 Volume Serial Number is 8093-8FED

 Directory of tmp

28.05.2014  15:37    <DIR>          .
28.05.2014  15:37    <DIR>          ..
28.05.2014  15:36    <DIR>          bin
28.05.2014  15:36               210 foo.scala
28.05.2014  15:37           506'336 test.js
               2 File(s)        506'546 bytes
               3 Dir(s)   1'862'295'552 bytes free

tmp>scalajsc foo.scala
foo.scala:1: error: object scalajs is not a member of package scala
import scala.scalajs.js.JSApp
             ^
foo.scala:2: error: not found: type JSApp
          object Foo extends JSApp {
                             ^
two errors found

tmp>scalajsc foo.scala

>

[sjrd]

Rescheduling this for 0.5.1. It's really not important. We can tell the very small intersection of developers using the CLI and Windows that it might fail on the first invocation in Known Issues.

[gzm0]

Note that his also happens if scalajsc is in the user path upon startup of cmd.exe.

[gzm0]

Could someone please test this independently, I'm running this in a VM and it wouldn't astonish me at all if it is a file share issue or something.

[sjrd]

I'll do that tonight if I don't forget.

[sjrd]

Well, seems it behaves exactly the same way on perfectly genuine, stand-alone Windows 8.1 installation.

[gzm0]

:( any windows gurus around?

[gzm0]

@sjrd I propose we move this to 0.5.2 so we get 0.5.1 free to release.

[sjrd]

Yes ok

[gzm0]

Putting on hold until someone actually complains