Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret ---+ Reporter: heisenbug | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.8.1 Component: Compiler| Version: 7.7 Keywords: | Os: Linux Architecture: x86_64 (amd64) | Failure: Runtime crash Difficulty: Unknown |Testcase: yes Blockedby: |Blocking: Related: | ---+ Comment(by heisenbug): It turns out that simply mixing GADTs + unsafeCoerce triggers the bug. (see freshly attached.. crash.hs) The order of the GADT constructors (being defined) seems to matter, swapping them changes a crash to some (apparently) undefined behaviour (neither output, nor pattern match failure). I suspect that unsafeCoerce does not preserve pointer tagging. -- Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler ___ Glasgow-haskell-bugs mailing list Glasgow-haskell-bugs@haskell.org http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret
+---
Reporter: heisenbug | Owner:
Type: bug| Status: closed
Priority: normal | Milestone: 7.8.1
Component: Compiler |Version: 7.7
Resolution: invalid| Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Runtime crash | Difficulty: Unknown
Testcase: yes| Blockedby:
Blocking: |Related:
+---
Changes (by igloo):
* status: new = closed
* resolution: = invalid
Comment:
`int-e` in #ghc diagnosed this:
{{{
GHC correctly infers that when it gets a value
of type forall a. TypedPair a, then there is no constructor
that can match.
And it happily creates a case statement without alternatives,
relying on the value to be bottom. The code, however, produces
a non-bottom value of the type. What will happen then is
unpredictable. (Interestingly, ghc 7.4.1 created a fresh
bottom value instead, which is arguably more friendly.)
}}}
{{{
Main.main :: GHC.Types.IO ()
[GblId, Str=DmdType b]
Main.main =
case Unsafe.Coerce.unsafeCoerce
@ (Main.TypedPair GHC.Types.Int)
@ (Main.TypedPair (GHC.Prim.Any *))
Main.$WTPInt
of _ {
}
}}}
This therefore looks like it isn't a bug - just an invalid use of
`unsafeCoerce`.
Thanks for taking the time to make such a small testcase, though. And
thanks too to `int-e` for the diagnosis.
--
Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:6
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler
___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret
+---
Reporter: heisenbug | Owner:
Type: bug| Status: closed
Priority: normal | Milestone: 7.8.1
Component: Compiler |Version: 7.7
Resolution: invalid| Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Runtime crash | Difficulty: Unknown
Testcase: yes| Blockedby:
Blocking: |Related:
+---
Comment(by int-e):
Interestingly, ghc 7.4.1 used to produce a {{{runTimeError Impossible
case alternative}}} for such cases rather than relying on the given value
to be bottom. Is this the result of an optimization (getting rid of
unreachable default branches in case, say) or an actual regression?
--
Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:7
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler
___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret ---+ Reporter: heisenbug | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.8.1 Component: Compiler| Version: 7.7 Keywords: | Os: Linux Architecture: x86_64 (amd64) | Failure: Runtime crash Difficulty: Unknown |Testcase: yes Blockedby: |Blocking: Related: | ---+ Changes (by igloo): * difficulty: = Unknown * milestone: = 7.8.1 -- Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:4 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler ___ Glasgow-haskell-bugs mailing list Glasgow-haskell-bugs@haskell.org http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret
---+
Reporter: heisenbug | Owner:
Type: bug| Status: new
Priority: normal | Component: Compiler
Version: 7.7| Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Runtime crash | Testcase: yes
Blockedby: | Blocking:
Related: |
---+
Changes (by heisenbug):
* cc: ggreif@… (added)
Comment:
I found a pre-7.4 compiler: ghc-7.4.0.20111220
it emits:
`internal-error: Impossible case alternative`
The other compiler did not even typecheck the test case:
{{{
ghc-7.5.20120713 internal-error.lhs
[1 of 1] Compiling Main ( internal-error.lhs, internal-error.o
)
internal-error.lhs:39:23:
Couldn't match type `BL.ByteString'
with
`bytestring-0.10.0.0:Data.ByteString.Lazy.Internal.ByteString'
Expected type: [ByteString]
-
bytestring-0.10.0.0:Data.ByteString.Lazy.Internal.ByteString
Actual type: [ByteString] - BL.ByteString
In the second argument of `(.)', namely `BL.fromChunks'
In the expression: decode . BL.fromChunks
In an equation for `decodeAs': decodeAs = decode . BL.fromChunks
internal-error.lhs:45:41:
Couldn't match expected type `BL.ByteString'
with actual type
`bytestring-0.10.0.0:Data.ByteString.Lazy.Internal.ByteString'
In the first argument of `BL.toChunks', namely `lazy'
In the second argument of `($)', namely `BL.toChunks lazy'
In the expression: decodeAs $ BL.toChunks lazy
}}}
--
Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:1
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler
___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret
---+
Reporter: heisenbug | Owner:
Type: bug| Status: new
Priority: normal | Component: Compiler
Version: 7.7| Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Runtime crash | Testcase: yes
Blockedby: | Blocking:
Related: |
---+
Comment(by heisenbug):
This is what gdb has to say:
{{{
(gdb) run
Starting program: /home/ggreif/ReleaseFeatureMatrix/internal-error
warning: no loadable sections found in added symbol-file system-supplied
DSO at 0x2aaab000
[Thread debugging using libthread_db enabled]
internal-error: internal error: stg_ap_v_ret
(GHC version 7.7.20121012 for x86_64_unknown_linux)
Please report this as a GHC bug:
http://www.haskell.org/ghc/reportabug
Program received signal SIGABRT, Aborted.
0x0035b2a30285 in raise () from /lib64/libc.so.6
(gdb) bt
#0 0x0035b2a30285 in raise () from /lib64/libc.so.6
#1 0x0035b2a31d30 in abort () from /lib64/libc.so.6
#2 0x2c623799 in rtsFatalInternalErrorFn (s=0x2c660128
stg_ap_v_ret,
ap=0x7fff8680) at rts/RtsMessages.c:170
#3 0x2c623304 in barf (s=0x2c660128 stg_ap_v_ret) at
rts/RtsMessages.c:42
#4 0x2c64d7ae in stg_ap_v_info ()
from /home/ggreif/lib/ghc-7.7.20121012/rts-1.0/libHSrts_debug-
ghc7.7.20121012.so
#5 0x in ?? ()
}}}
--
Ticket URL: http://hackage.haskell.org/trac/ghc/ticket/7323#comment:2
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler
___
Glasgow-haskell-bugs mailing list
Glasgow-haskell-bugs@haskell.org
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
Re: [GHC] #7323: decoding GADTs gives internal error: stg_ap_v_ret
#7323: decoding GADTs gives internal error: stg_ap_v_ret
---+
Reporter: heisenbug | Owner:
Type: bug| Status: new
Priority: normal | Component: Compiler
Version: 7.7| Keywords:
Os: Linux | Architecture: x86_64 (amd64)
Failure: Runtime crash | Testcase: yes
Blockedby: | Blocking:
Related: |
---+
Comment(by heisenbug):
Debug session on OS X (i386)
{{{
Breakpoint 1, 0x003bfe68 in stg_ap_v_info ()
(gdb) info reg
eax0x2808 10248 info table:
Main_TPString_con_info
ecx0x4 4 INFO TYPE (dito)
edx0xd00040 13631552
ebx0x425f4c 4349772
esp0xbfffd87c 0xbfffd87c
ebp0xd033ac 0xd033ac
esi0xd058f8 13654264
edi0xd05900 13654272
eip0x3bfe68 0x3bfe68 stg_ap_v_info+676
eflags 0x202514
cs 0x1b 27
ss 0x23 35
ds 0x23 35
es 0x23 35
fs 0x0 0
gs 0xf 15
(gdb) x/20x $eax-4
0x2804 Main_TPString_con_info_dsp+8: 0x00010004 0xff02c683
0x000c0065 0x9066
0x2814 Main_TPString_static_info_dsp: 0x003c216c 0x0002
0x00010007 0xff02c683
0x2824 Main_TPString_static_info+4: 0x000c0065 0x8955
0xec8353e5 0xe834
0x2834 main+10: 0x8d5b 0x3e380283 0x8b008b00
0x008b0450
0x2844 main+26: 0x89f04589 0x45c7f455 0x02f0
0xf0458b00
(gdb) x/20x $eax
0x2808 Main_TPString_con_info:0xff02c683 0x000c0065
0x9066 0x003c216c
0x2818 Main_TPString_static_info_dsp+4: 0x0002 0x00010007
0xff02c683 0x000c0065
0x2828 Main_TPString_static_info+8: 0x8955 0xec8353e5
0xe834 0x8d5b
0x2838 main+14: 0x3e380283 0x8b008b00 0x008b0450
0x89f04589
0x2848 main+30: 0x45c7f455 0x02f0 0xf0458b00
0x89f4558b
(gdb) p *(short*)0x2804
$1 = 4
}}}
It loads the address `Main_TPString_con_info` at this code position:
`Main_zdWTPString_info+0x14`
{{{
_Main_zdWTPString_info_dsp:
0044orb $_Main_zdWTPInt_info_dsp,%al
0046addb(%eax),%al
0048addb%al,(%eax)
004aaddb%al,(%eax)
004csldt(%eax)
004faddb%al,0x7b3b0cc7(%ebx)
0055popl%esp
0056ja 0x0073
0058movl$_Main_TPString_con_info,0xf8(%edi)
005fmovl_Main_zdWTPInt_info_dsp(%ebp),%eax
0062movl%eax,0xfc(%edi)
0065movl0x04(%ebp),%eax
0068movl%eax,(%edi)
006aleal0xfa(%edi),%esi
006daddl$0x08,%ebp
0070jmp *_Main_zdWTPInt_info_dsp(%ebp) calls
stg_upd_frame_info
0073movl$_Main_zdWTPInt_info,0x74(%ebx)
007amovl$_Main_zdWTPString_closure,%esi
007fjmp *0xfc(%ebx)
0082orb $_Main_zdWTPInt_info_dsp,%al
0084addb%al,(%eax)
0086nop
}}}
{{{
0x003bfa18 stg_upd_frame_info+0: mov0x4(%ebp),%eax
0x003bfa1b stg_upd_frame_info+3: add$0x8,%ebp
0x003bfa1e stg_upd_frame_info+6: sub$0x8,%esp
0x003bfa21 stg_upd_frame_info+9: push %eax
0x003bfa22 stg_upd_frame_info+10: mov%eax,0x4c(%esp)
0x003bfa26 stg_upd_frame_info+14: call 0x3892d5
overwritingClosure overwritingClosure (p=0xd058e8) at
ClosureMacros.h:474
0x003bfa2b stg_upd_frame_info+19: add$0xc,%esp
0x003bfa2e stg_upd_frame_info+22: mov0x40(%esp),%eax
0x003bfa32 stg_upd_frame_info+26: mov%esi,0x4(%eax)
0x003bfa35 stg_upd_frame_info+29: mov0x3e63d8,%ecx
0x003bfa3b stg_upd_frame_info+35: mov%ecx,(%eax)
0x003bfa3d stg_upd_frame_info+37: mov%eax,%ecx
0x003bfa3f stg_upd_frame_info+39: and$0xfff0,%ecx
0x003bfa45 stg_upd_frame_info+45: mov%eax,%edx
0x003bfa47 stg_upd_frame_info+47: and$0xff000,%edx
0x003bfa4d stg_upd_frame_info+53: shr$0x7,%edx
0x003bfa50 stg_upd_frame_info+56: or %ecx,%edx
0x003bfa52 stg_upd_frame_info+58: cmpw $0x0,0x14(%edx)
0x003bfa57 stg_upd_frame_info+63: jne0x3bfa5c
stg_upd_frame_info+68
0x003bfa59 stg_upd_frame_info+65: jmp*0x0(%ebp)
stg_marked_upd_frame_info
0x003bfa5c stg_upd_frame_info+68: movswl 0x14(%edx),%ecx
0x003bfa60 stg_upd_frame_info+72: shl$0x2,%ecx
0x003bfa63 stg_upd_frame_info+75: mov0x9c(%ebx),%edx
0x003bfa69 stg_upd_frame_info+81: add
