
Do you have access to an AMD64 machine? I pulled from yhc-devel and the build broke in the same place. This time the offending definition and declaration were: IntegerNode* decodeFloat64(Float64 f, Int32* exp); IntegerNode* decodeFloat64(Float64 f, Int* rExp) The patches attached. Thanks, Michal On Wed, 2005-11-23 at 16:32 +0000, Neil Mitchell wrote:
Hi Michał ,
We're still working on fixing your bug, but you might be interested to know there is now a yhc mailing list: http://haskell.org/mailman/listinfo/yhc
If its still breaking in another week start prodding at that list!
Thanks
Neil
On 11/20/05, Michał Pałka
wrote: Hello,
I would like to report a bug of YHC. Here is the description.
What I did: 1) darcs get http://www.cs.york.ac.uk/fp/darcs/yhc 2) chmod u+x configure; ./configure 3) make 4) The build broke, so I made a change that is attached and finished the build. 5) I compiled using yhc a hello world module (main = putStrLn "Hello") 6) and tried to run it with yhi and received a segv. 7) After changing the runtime Makefile I compiled it with -g and got this stack trace: Starting program: /home/michal/prg/haskell/yhc/inst/bin/yhi Main
Program received signal SIGSEGV, Segmentation fault. 0x000000000040ab10 in run (top=0x2aaaab1c60b0) at mutator.c:79 79 Switch (gdb) bt #0 0x000000000040ab10 in run (top=0x2aaaab1c60b0) at mutator.c:79 #1 0x00000000004041f7 in main (argc=2, argv=0x7fffffa132f8) at main.c:113
I also got a segfault with the devel branch, but wasn't able to compile it with -g. And it broke when building libraries.
Let me know if you need more information.
Thanks, Michal

Michał Pałka wrote:
Do you have access to an AMD64 machine?
I don't unfortunately. It *used* to work on AMD64 at one point, but it should really be tested regularly. Unfortunately debugging the runtime system requires quite a bit of knowledge about the inner workings of YHC. I'll try and get it tested on someone elses AMD64 machine soon.
I pulled from yhc-devel and the build broke in the same place. This time the offending definition and declaration were: IntegerNode* decodeFloat64(Float64 f, Int32* exp); IntegerNode* decodeFloat64(Float64 f, Int* rExp)
Right, hopefully I've fixed those now. It should have been 'Int' rather than Int32 because it ends up being put in a Haskell Int (which is a machine sized word). thanks Tom

A fix for decodeFloat32 needed on my system. On Thu, 2005-11-24 at 12:43 +0000, Thomas Shackell wrote:
Michał Pałka wrote:
Do you have access to an AMD64 machine?
I don't unfortunately. It *used* to work on AMD64 at one point, but it should really be tested regularly. Unfortunately debugging the runtime system requires quite a bit of knowledge about the inner workings of YHC.
I'll try and get it tested on someone elses AMD64 machine soon.
I pulled from yhc-devel and the build broke in the same place. This time the offending definition and declaration were: IntegerNode* decodeFloat64(Float64 f, Int32* exp); IntegerNode* decodeFloat64(Float64 f, Int* rExp)
Right, hopefully I've fixed those now. It should have been 'Int' rather than Int32 because it ends up being put in a Haskell Int (which is a machine sized word).
thanks
Tom

Done. Tom Michał Pałka wrote:
A fix for decodeFloat32 needed on my system.
On Thu, 2005-11-24 at 12:43 +0000, Thomas Shackell wrote:
Michał Pałka wrote:
Do you have access to an AMD64 machine?
I don't unfortunately. It *used* to work on AMD64 at one point, but it should really be tested regularly. Unfortunately debugging the runtime system requires quite a bit of knowledge about the inner workings of YHC.
I'll try and get it tested on someone elses AMD64 machine soon.
I pulled from yhc-devel and the build broke in the same place. This time the offending definition and declaration were: IntegerNode* decodeFloat64(Float64 f, Int32* exp); IntegerNode* decodeFloat64(Float64 f, Int* rExp)
Right, hopefully I've fixed those now. It should have been 'Int' rather than Int32 because it ends up being put in a Haskell Int (which is a machine sized word).
thanks
Tom
------------------------------------------------------------------------
New patches:
[Fixed arg types of decodeFloat32 to match declaration on AMD64 Michal Palka
**20051124140508] { hunk ./src/runtime/BCKernel/integer.c 112 -IntegerNode* decodeFloat32(Float32 f, Int32* rExp){ +IntegerNode* decodeFloat32(Float32 f, Int* rExp){ } Context:
[Fixed some AM64 issues (integer.c/integer.h) Tom Shackell
**20051124123931] [TAG Stable 1 Tom Shackell **20051123104200] Patch bundle hash: 3452c7808756ee6424c96c065df46fce8bd7fff7
participants (2)
-
Michał Pałka
-
Thomas Shackell