Hi all,

I've found that if I run

darcs record

and then hit ctrl-C when prompted for the patch name I get the
following error:

Shall I record this change? (1/8)  [ynWsfvpxdaqjk], or ? for help: a
What is the patch name? foodarcs: main thread exited (uncaught exception)

and the lock file is left in _darcs/lock.  This is a pretty serious
regression, and I suspect it relates to the threading code that was
added to deal with I don't recall what silliness we ran into when we
started using ghc with -threaded.  Anyhow, we need to look closely at
that code.

David

On Mon, Sep 22, 2008 at 6:37 PM, David Roundy <bugs@darcs.net> wrote:
>
> New submission from David Roundy <droundy@darcs.net>:
>
> Hi all,
>
> I've found that if I run
>
> darcs record
>
> and then hit ctrl-C when prompted for the patch name I get the
> following error:
>
> Shall I record this change? (1/8)  [ynWsfvpxdaqjk], or ? for help: a
> What is the patch name? foodarcs: main thread exited (uncaught exception)
>
> and the lock file is left in _darcs/lock.  This is a pretty serious
> regression, and I suspect it relates to the threading code that was
> added to deal with I don't recall what silliness we ran into when we
> started using ghc with -threaded.  Anyhow, we need to look closely at
> that code.

To help narrow the search space, I just tried your example with:
$ darcs --version
2.0.2 (+ 144 patches)

and I get the usual:
Shall I record this change? (2/6)  [ynWsfvpxdaqjk], or ? for help: d
What is the patch name? blah^CwithSignalsHandled: Interrupted!

This is on OSX 10.5.

Jason

On Mon, Sep 22, 2008 at 19:00:33 -0700, Jason Dagit wrote:
> To help narrow the search space, I just tried your example with:
> $ darcs --version
> 2.0.2 (+ 144 patches)
> 
> and I get the usual:
> Shall I record this change? (2/6)  [ynWsfvpxdaqjk], or ? for help: d
> What is the patch name? blah^CwithSignalsHandled: Interrupted!
> 
> This is on OSX 10.5.

Ah, but this works for me too even with unstable
  darcs 2.0.3pre1 (+ 41 patches)

Maybe it's specific to Linux

On Tue, Sep 23, 2008 at 09:20:38AM +0100, Eric Kow wrote:
> On Mon, Sep 22, 2008 at 19:00:33 -0700, Jason Dagit wrote:
> > To help narrow the search space, I just tried your example with:
> > $ darcs --version
> > 2.0.2 (+ 144 patches)
> > 
> > and I get the usual:
> > Shall I record this change? (2/6)  [ynWsfvpxdaqjk], or ? for help: d
> > What is the patch name? blah^CwithSignalsHandled: Interrupted!
> > 
> > This is on OSX 10.5.
> 
> Ah, but this works for me too even with unstable
>   darcs 2.0.3pre1 (+ 41 patches)
> 
> Maybe it's specific to Linux

It turns out to triggered by my Workaround refactors that I haven't
yet pushed.  I'm still tracking it down...

David

Okay, it turns out that in removing the Workaround, I'd accidentally removed the
check to see if we can use System.Posix.Signals... I just checked if
HAVE_SIGNALS was defined without ever defining it.  So this is now fixed.  Silly me.

David