- About Scala
- Documentation
- Code Examples
- Software
- Scala Developers
Which has newer library, 2.8.0prebeta1, or Eclipse plugin 2.8.0.r21030-b20100302035619
Wed, 2010-03-03, 19:32
I ask, because our gigantic application (Fortress, in this case, the Fortress interpreter) dies (on a few inputs) when batch-compiled against 2.8.0prebeta1, but runs fine when refreshed-cleaned-built within Eclipse.
So either there is an old bug that has been fixed, or a new bug waiting to appear in the next crank-turn of the Scala release.
Also, can anyone provide any advice for batch-compiling using the plugin jar files?
I wanted to do an apples-to-apples comparison, just in case it was instead triggered by the mode used to invoke the compiler.
I tried what I though was obvious
but it failed with
"taskdef A class needed by class scala.tools.ant.FastScalac cannot be found: scala/ScalaObject"
which I believe means I have goofed up my path.
Any help is appreciated, the goal of this exercise (2.8.0 upgrade) was to get access to the shiny new Eclipse plugin to better be able to debug code, but the bugs are not supposed to disappear when I look for them in Eclipse. Grrrrr.
David
Wed, 2010-03-03, 20:17
#2
Re: Which has newer library, 2.8.0prebeta1, or Eclipse plugin 2
On 2010-03-03, at 2:01 PM, Miles Sabin wrote:
> In this case, given that you're seeing the bug on
> 2.8.0prebeta1 but not on trunk, you probably ought to start working
> with nightly builds of the command line tools as well as of Eclipse.
I ought, if it were just me, but there's more than just me working on this.
Any idea when prebeta2 will show up?
Till then, I've got to figure out how to whack our code to work around this bug.
There's also a possibility that this is just "implementation-dependent behavior"
that our code happens to depend on; we hit that with Stack.dropWhile.
David
Thu, 2010-03-04, 00:07
#3
Re: Which has newer library, 2.8.0prebeta1, or Eclipse plugin 2
FYI, the difference in behavior, was caused by two things:
1) increased stack consumption in the newer versions of Scala, which caused some stack overflow failures
and
2) different (and tricky to sort out) stack size rules within Eclipse debug sessions, junit processes forked by ant, and command-line invocations of the program.
Once that was all sorted out, everything behaved properly. There is no bug. (Or rather, what I observed, was not a Scala bug; I should not rule out either PEBKAC or other bugs in Scala.)
David
On Wed, Mar 3, 2010 at 6:32 PM, David Chase wrote:
> I ask, because our gigantic application (Fortress, in this case, the
> Fortress interpreter) dies (on a few inputs) when batch-compiled against
> 2.8.0prebeta1, but runs fine when refreshed-cleaned-built within Eclipse.
The clue is in the version number you quoted,
Eclipse plugin 2.8.0.r21030-b20100302035619
^^^^^ ^^^^^^^^
This is the nightly build of the plugin of the 2nd of March, ie. the
most recent nightly build. It includes the corresponding nightly of
the main Scala toolchain, so it's quite a bit more recent than
2.8.0prebeta1.
You should always use matched versions of Eclipse and the Scala
command line tools. In this case, given that you're seeing the bug on
2.8.0prebeta1 but not on trunk, you probably ought to start working
with nightly builds of the command line tools as well as of Eclipse.
Cheers,
Miles