RE: Building on Windows: the saga continues

I've opened a ticket for this: http://hackage.haskell.org/trac/ghc/ticket/7510

Simon

| -----Original Message-----
| From: [email protected] [mailto:[email protected]]
| On Behalf Of Ian Lynagh
| Sent: 18 December 2012 18:22
| To: [email protected]
| Subject: Re: Building on Windows: the saga continues
| 
| On Mon, Dec 17, 2012 at 02:01:28PM +0000, Simon Peyton-Jones wrote:
| > | On Mon, Dec 17, 2012 at 09:21:44AM +0000, Simon Peyton-Jones wrote:
| > | >
| > | > o   There's a seg-fault
| > |
| > | I can reproduce this. I'll look into it.
| >
| > Great, thank you.
| >
| > It turns out that simply
| >
| > 	ghc-stage2 -debug
| > or
| > 	ghc-stage2 -prof
| >
| > is enough to trigger the seg-fault.  Any "way" flag seems to do it.
| 
| Unfortunately, I've run out of time to debug this for now. Here's what
| I've got:
| 
| It seems that the commits
| 
|     0c4a9f38637dfc3bc8fd48e8ba6bf64da51b727b
|     ecd967612877e1965ddebefe9b83acd837bb413a
| 
| introduced the problem, but I'm not sure whether they are the direct
| cause or just happened to tickle a pre-existing bug. Attached is a self-
| contained module that goes wrong:
| 
| $ inplace/bin/ghc-stage1 -debug --make test -O -fforce-recomp $ ./test
| +RTS -DS; echo $?
| cap 0: initialised
| A
| Segmentation fault/access violation in generated code
| 1
| 
| Works with 7.4.1:
| $ ghc -debug --make test -O -fforce-recomp $ ./test; echo $?
| A
| [WayDyn]
| B
| 0
| 
| It's quite fragile, e.g. if I remove
|     ghcMode               :: (),
| from the DynFlags type then it succeeds.
| 
| 
| Thanks
| Ian