Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.
Sign upCompiler using HARP + MPS fails to compile common-dylan-test-suite-app #882
Comments
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
|
cc: @rptb1 @gareth-rees |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
Me either. I was thinking that it might have been one of two things:
- Is it possible that it was due to touching the memory that contains the bad floating point value rather than the act of carrying out an operation on it?
- Is there any way that the delivery of a
SIGFPEcan be delayed?
This is pretty easy to reproduce at least (just takes some time) if you hop on IRC and need help. Otherwise, I'm wondering if I should try to fix the build so that MPS has better debug data / symbols available and see what I see them from the MPS side of things.
I'm not convinced that it is necessarily MPS at fault, just that the symptom lies there (in the asserts).
|
Me either. I was thinking that it might have been one of two things:
This is pretty easy to reproduce at least (just takes some time) if you hop on IRC and need help. Otherwise, I'm wondering if I should try to fix the build so that MPS has better debug data / symbols available and see what I see them from the MPS side of things. I'm not convinced that it is necessarily MPS at fault, just that the symptom lies there (in the asserts). |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
gareth-rees
Apr 16, 2015
Yes, if you can find out which floating-point operation in TraceStart is invalid, and what the arguments are, that would be very useful.
gareth-rees
commented
Apr 16, 2015
|
Yes, if you can find out which floating-point operation in |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
Okay, you asked for it.
I have a new theory: Perhaps with the HARP run-time, we aren't resetting the FPU status correctly after handling an error and that's causing something in MPS using floating point to think it had an error. This is pre-<fenv.h> code, so it is a bit hairy.
0x4024af92 in TraceStart ()
from /home/bruce/Development/opendylan/Bootstrap.3/bin/../lib/libdylan.so
(gdb) disassemble
Dump of assembler code for function TraceStart:
0x4024ad8d <+0>: push %ebp
0x4024ad8e <+1>: push %edi
0x4024ad8f <+2>: push %esi
0x4024ad90 <+3>: push %ebx
0x4024ad91 <+4>: sub $0x4c,%esp
0x4024ad94 <+7>: mov 0x60(%esp),%ebp
0x4024ad98 <+11>: fldl 0x64(%esp)
0x4024ad9c <+15>: fstpl 0x10(%esp)
0x4024ada0 <+19>: fldl 0x6c(%esp)
0x4024ada4 <+23>: fstpl 0x18(%esp)
0x4024ada8 <+27>: mov %ebp,(%esp)
0x4024adab <+30>: call 0x40215afa <TraceCheck>
0x4024adb0 <+35>: test %eax,%eax
0x4024adb2 <+37>: jne 0x4024add0 <TraceStart+67>
0x4024adb4 <+39>: movl $0x40262b80,0x8(%esp)
0x4024adbc <+47>: movl $0x658,0x4(%esp)
0x4024adc4 <+55>: movl $0x40259c80,(%esp)
0x4024adcb <+62>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024add0 <+67>: cmpl $0x1,0x18(%ebp)
0x4024add4 <+71>: je 0x4024adf2 <TraceStart+101>
0x4024add6 <+73>: movl $0x40264d37,0x8(%esp)
0x4024adde <+81>: movl $0x659,0x4(%esp)
0x4024ade6 <+89>: movl $0x40259c80,(%esp)
0x4024aded <+96>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024adf2 <+101>: fldz
0x4024adf4 <+103>: fldl 0x10(%esp)
0x4024adf8 <+107>: fucomip %st(1),%st
0x4024adfa <+109>: fstp %st(0)
0x4024adfc <+111>: jae 0x4024ae1a <TraceStart+141>
0x4024adfe <+113>: movl $0x402661db,0x8(%esp)
0x4024ae06 <+121>: movl $0x65a,0x4(%esp)
0x4024ae0e <+129>: movl $0x40259c80,(%esp)
0x4024ae15 <+136>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae1a <+141>: fld1
0x4024ae1c <+143>: fldl 0x10(%esp)
0x4024ae20 <+147>: fxch %st(1)
0x4024ae22 <+149>: fucomip %st(1),%st
0x4024ae24 <+151>: fstp %st(0)
0x4024ae26 <+153>: jae 0x4024ae44 <TraceStart+183>
0x4024ae28 <+155>: movl $0x402661ec,0x8(%esp)
0x4024ae30 <+163>: movl $0x65b,0x4(%esp)
0x4024ae38 <+171>: movl $0x40259c80,(%esp)
0x4024ae3f <+178>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae44 <+183>: fldz
0x4024ae46 <+185>: fldl 0x18(%esp)
0x4024ae4a <+189>: fucomip %st(1),%st
0x4024ae4c <+191>: fstp %st(0)
0x4024ae4e <+193>: jae 0x4024ae6c <TraceStart+223>
0x4024ae50 <+195>: movl $0x402661fd,0x8(%esp)
0x4024ae58 <+203>: movl $0x65c,0x4(%esp)
0x4024ae60 <+211>: movl $0x40259c80,(%esp)
0x4024ae67 <+218>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae6c <+223>: mov 0x8(%ebp),%ebx
0x4024ae6f <+226>: mov %ebx,0x4(%esp)
0x4024ae73 <+230>: lea 0x30(%esp),%eax
0x4024ae77 <+234>: mov %eax,(%esp)
0x4024ae7a <+237>: call 0x4021c5ef <SegFirst>
0x4024ae7f <+242>: test %eax,%eax
0x4024ae81 <+244>: je 0x4024af44 <TraceStart+439>
0x4024ae87 <+250>: mov 0x30(%esp),%eax
0x4024ae8b <+254>: mov 0x14(%eax),%edi
0x4024ae8e <+257>: mov 0x8(%eax),%edx
0x4024ae91 <+260>: mov 0x8(%edx),%esi
0x4024ae94 <+263>: movzbl 0x19(%eax),%edx
0x4024ae98 <+267>: and $0x1,%edx
0x4024ae9b <+270>: mov 0x4(%ebp),%eax
0x4024ae9e <+273>: bt %eax,%edx
0x4024aea1 <+276>: jae 0x4024aebf <TraceStart+306>
0x4024aea3 <+278>: movl $0x4025c4e4,0x8(%esp)
0x4024aeab <+286>: movl $0x66b,0x4(%esp)
0x4024aeb3 <+294>: movl $0x40259c80,(%esp)
0x4024aeba <+301>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024aebf <+306>: mov 0x30(%esp),%eax
0x4024aec3 <+310>: testb $0x78,0x19(%eax)
0x4024aec7 <+314>: je 0x4024af24 <TraceStart+407>
0x4024aec9 <+316>: sub %esi,%edi
0x4024aecb <+318>: mov 0x10(%ebp),%edx
0x4024aece <+321>: test %edx,0x24(%eax)
0x4024aed1 <+324>: je 0x4024aefe <TraceStart+369>
0x4024aed3 <+326>: mov %eax,0x8(%esp)
0x4024aed7 <+330>: mov %ebp,0x4(%esp)
0x4024aedb <+334>: mov 0x8(%eax),%eax
0x4024aede <+337>: mov (%eax),%eax
0x4024aee0 <+339>: mov %eax,(%esp)
0x4024aee3 <+342>: call 0x40216cef <PoolGrey>
0x4024aee8 <+347>: mov 0x30(%esp),%eax
0x4024aeec <+351>: movzbl 0x19(%eax),%edx
0x4024aef0 <+355>: and $0x1,%edx
0x4024aef3 <+358>: mov 0x4(%ebp),%eax
0x4024aef6 <+361>: bt %eax,%edx
0x4024aef9 <+364>: jae 0x4024aefe <TraceStart+369>
0x4024aefb <+366>: add %edi,0x3c(%ebp)
0x4024aefe <+369>: mov 0x30(%esp),%eax
0x4024af02 <+373>: mov 0x8(%eax),%edx
0x4024af05 <+376>: mov (%edx),%edx
0x4024af07 <+378>: mov 0x8(%edx),%edx
0x4024af0a <+381>: testb $0x2,0x1c(%edx)
0x4024af0e <+385>: je 0x4024af24 <TraceStart+407>
0x4024af10 <+387>: movzbl 0x19(%eax),%edx
0x4024af14 <+391>: shr %dl
0x4024af16 <+393>: and $0x1,%edx
0x4024af19 <+396>: mov 0x4(%ebp),%eax
0x4024af1c <+399>: bt %eax,%edx
0x4024af1f <+402>: jb 0x4024af24 <TraceStart+407>
0x4024af21 <+404>: add %edi,0x38(%ebp)
0x4024af24 <+407>: mov 0x30(%esp),%eax
0x4024af28 <+411>: mov %eax,0x8(%esp)
0x4024af2c <+415>: mov %ebx,0x4(%esp)
0x4024af30 <+419>: lea 0x30(%esp),%eax
0x4024af34 <+423>: mov %eax,(%esp)
0x4024af37 <+426>: call 0x4021c5c0 <SegNext>
0x4024af3c <+431>: test %eax,%eax
0x4024af3e <+433>: jne 0x4024ae87 <TraceStart+250>
0x4024af44 <+439>: mov %ebp,0x8(%esp)
0x4024af48 <+443>: movl $0x4021668b,0x4(%esp)
0x4024af50 <+451>: mov %ebx,(%esp)
0x4024af53 <+454>: call 0x4020a32c <RootsIterate>
0x4024af58 <+459>: test %eax,%eax
0x4024af5a <+461>: je 0x4024af78 <TraceStart+491>
0x4024af5c <+463>: movl $0x40266315,0x8(%esp)
0x4024af64 <+471>: movl $0x689,0x4(%esp)
0x4024af6c <+479>: movl $0x40259c80,(%esp)
0x4024af73 <+486>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024af78 <+491>: mov 0x34(%ebp),%eax
0x4024af7b <+494>: mov $0x0,%edx
0x4024af80 <+499>: mov %eax,0x20(%esp)
0x4024af84 <+503>: mov %edx,0x24(%esp)
0x4024af88 <+507>: fildll 0x20(%esp)
0x4024af8c <+511>: fld1
0x4024af8e <+513>: fldl 0x10(%esp)
=> 0x4024af92 <+517>: fsubr %st(1),%st
0x4024af94 <+519>: fmulp %st,%st(2)
0x4024af96 <+521>: fxch %st(1)
0x4024af98 <+523>: fnstcw 0x2e(%esp)
0x4024af9c <+527>: movzwl 0x2e(%esp),%eax
0x4024afa1 <+532>: mov $0xc,%ah
0x4024afa3 <+534>: mov %ax,0x2c(%esp)
0x4024afa8 <+539>: fldcw 0x2c(%esp)
0x4024afac <+543>: fistpll 0x20(%esp)
0x4024afb0 <+547>: fldcw 0x2e(%esp)
0x4024afb4 <+551>: mov 0x20(%esp),%eax
0x4024afb8 <+555>: fldl 0x18(%esp)
0x4024afbc <+559>: fmuls 0x402678a8
0x4024afc2 <+565>: fxch %st(1)
0x4024afc4 <+567>: fucomip %st(1),%st
0x4024afc6 <+569>: ja 0x4024afd8 <TraceStart+587>
0x4024afc8 <+571>: fldl 0x402678e8
0x4024afce <+577>: fxch %st(1)
0x4024afd0 <+579>: fucomi %st(1),%st
0x4024afd2 <+581>: fcmovnb %st(1),%st
0x4024afd4 <+583>: fstp %st(1)
0x4024afd6 <+585>: jmp 0x4024afdc <TraceStart+591>
0x4024afd8 <+587>: fstp %st(0)
0x4024afda <+589>: fld1
0x4024afdc <+591>: add 0x3c(%ebp),%eax
0x4024afdf <+594>: fldcw 0x2c(%esp)
0x4024afe3 <+598>: fistpll 0x20(%esp)
0x4024afe7 <+602>: fldcw 0x2e(%esp)
0x4024afeb <+606>: mov 0x20(%esp),%ecx
0x4024afef <+610>: mov $0x0,%edx
0x4024aff4 <+615>: div %ecx
0x4024aff6 <+617>: add $0x1,%eax
0x4024aff9 <+620>: mov %eax,0x40(%ebp)
0x4024affc <+623>: mov 0x403011a8,%eax
0x4024b001 <+628>: sub $0x402fc1a0,%eax
0x4024b006 <+633>: cmp $0x33,%eax
0x4024b009 <+636>: ja 0x4024b042 <TraceStart+693>
0x4024b00b <+638>: movl $0x2,(%esp)
0x4024b012 <+645>: call 0x4022d788 <EventFlush>
0x4024b017 <+650>: mov 0x403011a8,%eax
0x4024b01c <+655>: sub $0x402fc1a0,%eax
0x4024b021 <+660>: cmp $0x33,%eax
0x4024b024 <+663>: ja 0x4024b042 <TraceStart+693>
0x4024b026 <+665>: movl $0x4025ee34,0x8(%esp)
0x4024b02e <+673>: movl $0x6a2,0x4(%esp)
0x4024b036 <+681>: movl $0x40259c80,(%esp)
0x4024b03d <+688>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b042 <+693>: mov 0x403011a8,%ecx
0x4024b048 <+699>: movw $0x1c,-0x34(%ecx)
0x4024b04e <+705>: movw $0x34,-0x32(%ecx)
0x4024b054 <+711>: rdtsc
0x4024b056 <+713>: mov %eax,-0x30(%ecx)
0x4024b059 <+716>: mov %edx,-0x2c(%ecx)
0x4024b05c <+719>: mov %ebp,-0x28(%ecx)
0x4024b05f <+722>: fldl 0x10(%esp)
0x4024b063 <+726>: fstpl -0x24(%ecx)
0x4024b066 <+729>: fldl 0x18(%esp)
0x4024b06a <+733>: fstpl -0x1c(%ecx)
0x4024b06d <+736>: mov 0x34(%ebp),%eax
0x4024b070 <+739>: mov %eax,-0x14(%ecx)
0x4024b073 <+742>: mov 0x38(%ebp),%eax
0x4024b076 <+745>: mov %eax,-0x10(%ecx)
0x4024b079 <+748>: mov 0x3c(%ebp),%eax
0x4024b07c <+751>: mov %eax,-0xc(%ecx)
0x4024b07f <+754>: mov 0x10(%ebp),%eax
0x4024b082 <+757>: mov %eax,-0x8(%ecx)
0x4024b085 <+760>: mov 0x40(%ebp),%eax
0x4024b088 <+763>: mov %eax,-0x4(%ecx)
0x4024b08b <+766>: subl $0x34,0x403011a8
0x4024b092 <+773>: movl $0x2,0x18(%ebp)
0x4024b099 <+780>: mov %ebp,(%esp)
0x4024b09c <+783>: call 0x4022c6fb <TracePostStartMessage>
0x4024b0a1 <+788>: mov %ebp,(%esp)
0x4024b0a4 <+791>: call 0x40215afa <TraceCheck>
0x4024b0a9 <+796>: test %eax,%eax
0x4024b0ab <+798>: jne 0x4024b0c9 <TraceStart+828>
0x4024b0ad <+800>: movl $0x40262b80,0x8(%esp)
0x4024b0b5 <+808>: movl $0x233,0x4(%esp)
0x4024b0bd <+816>: movl $0x40259c80,(%esp)
0x4024b0c4 <+823>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b0c9 <+828>: mov 0x4(%ebp),%ecx
0x4024b0cc <+831>: mov $0x1,%eax
0x4024b0d1 <+836>: shl %cl,%eax
0x4024b0d3 <+838>: mov %eax,0x34(%esp)
0x4024b0d7 <+842>: mov 0x8(%ebp),%eax
0x4024b0da <+845>: mov %eax,0x18(%esp)
0x4024b0de <+849>: mov %eax,0x38(%esp)
0x4024b0e2 <+853>: mov %eax,(%esp)
0x4024b0e5 <+856>: call 0x4024384c <ShieldSuspend>
0x4024b0ea <+861>: cmpl $0x2,0x18(%ebp)
0x4024b0ee <+865>: je 0x4024b10c <TraceStart+895>
0x4024b0f0 <+867>: movl $0x4025bd24,0x8(%esp)
0x4024b0f8 <+875>: movl $0x23a,0x4(%esp)
0x4024b100 <+883>: movl $0x40259c80,(%esp)
0x4024b107 <+890>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b10c <+895>: mov 0x4(%ebp),%eax
0x4024b10f <+898>: mov 0x18(%esp),%ecx
0x4024b113 <+902>: mov 0x41c(%ecx),%edx
0x4024b119 <+908>: bt %eax,%edx
0x4024b11c <+911>: jae 0x4024b13a <TraceStart+941>
0x4024b11e <+913>: movl $0x4025ee8c,0x8(%esp)
0x4024b126 <+921>: movl $0x23b,0x4(%esp)
0x4024b12e <+929>: movl $0x40259c80,(%esp)
0x4024b135 <+936>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b13a <+941>: mov 0x403011a8,%eax
0x4024b13f <+946>: sub $0x402fc1a0,%eax
0x4024b144 <+951>: cmp $0x13,%eax
0x4024b147 <+954>: ja 0x4024b180 <TraceStart+1011>
0x4024b149 <+956>: movl $0x2,(%esp)
0x4024b150 <+963>: call 0x4022d788 <EventFlush>
0x4024b155 <+968>: mov 0x403011a8,%eax
0x4024b15a <+973>: sub $0x402fc1a0,%eax
0x4024b15f <+978>: cmp $0x13,%eax
0x4024b162 <+981>: ja 0x4024b180 <TraceStart+1011>
0x4024b164 <+983>: movl $0x4025eebc,0x8(%esp)
0x4024b16c <+991>: movl $0x23d,0x4(%esp)
0x4024b174 <+999>: movl $0x40259c80,(%esp)
0x4024b17b <+1006>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b180 <+1011>: mov 0x403011a8,%ecx
0x4024b186 <+1017>: movw $0x20,-0x14(%ecx)
0x4024b18c <+1023>: movw $0x14,-0x12(%ecx)
0x4024b192 <+1029>: rdtsc
0x4024b194 <+1031>: mov %eax,-0x10(%ecx)
0x4024b197 <+1034>: mov %edx,-0xc(%ecx)
0x4024b19a <+1037>: mov %ebp,-0x8(%ecx)
0x4024b19d <+1040>: mov 0x18(%esp),%eax
0x4024b1a1 <+1044>: mov %eax,-0x4(%ecx)
0x4024b1a4 <+1047>: subl $0x14,0x403011a8
0x4024b1ab <+1054>: mov 0x50(%eax),%esi
0x4024b1ae <+1057>: mov (%esi),%edi
0x4024b1b0 <+1059>: add $0x50,%eax
0x4024b1b3 <+1062>: cmp %eax,%esi
0x4024b1b5 <+1064>: je 0x4024b21b <TraceStart+1166>
0x4024b1b7 <+1066>: mov %ebp,0x60(%esp)
0x4024b1bb <+1070>: mov %eax,%ebp
0x4024b1bd <+1072>: lea -0x10(%esi),%eax
0x4024b1c0 <+1075>: mov %eax,(%esp)
0x4024b1c3 <+1078>: call 0x40214d34 <PoolCheck>
0x4024b1c8 <+1083>: test %eax,%eax
0x4024b1ca <+1085>: jne 0x4024b1e8 <TraceStart+1115>
0x4024b1cc <+1087>: movl $0x40262a05,0x8(%esp)
0x4024b1d4 <+1095>: movl $0x1c0,0x4(%esp)
0x4024b1dc <+1103>: movl $0x40259c80,(%esp)
0x4024b1e3 <+1110>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b1e8 <+1115>: mov 0x8(%esi),%eax
0x4024b1eb <+1118>: mov (%eax),%ebx
0x4024b1ed <+1120>: add $0x8,%esi
0x4024b1f0 <+1123>: cmp %esi,%eax
0x4024b1f2 <+1125>: je 0x4024b20b <TraceStart+1150>
0x4024b1f4 <+1127>: sub $0x14,%eax
0x4024b1f7 <+1130>: mov %eax,(%esp)
0x4024b1fa <+1133>: call 0x4021916f <BufferFlip>
0x4024b1ff <+1138>: mov (%ebx),%edx
0x4024b201 <+1140>: cmp %ebx,%esi
0x4024b203 <+1142>: je 0x4024b20b <TraceStart+1150>
0x4024b205 <+1144>: mov %ebx,%eax
0x4024b207 <+1146>: mov %edx,%ebx
0x4024b209 <+1148>: jmp 0x4024b1f4 <TraceStart+1127>
0x4024b20b <+1150>: mov (%edi),%eax
0x4024b20d <+1152>: cmp %edi,%ebp
0x4024b20f <+1154>: je 0x4024b217 <TraceStart+1162>
0x4024b211 <+1156>: mov %edi,%esi
0x4024b213 <+1158>: mov %eax,%edi
0x4024b215 <+1160>: jmp 0x4024b1bd <TraceStart+1072>
0x4024b217 <+1162>: mov 0x60(%esp),%ebp
0x4024b21b <+1166>: mov 0x14(%ebp),%eax
0x4024b21e <+1169>: test %eax,%eax
0x4024b220 <+1171>: je 0x4024b357 <TraceStart+1482>
0x4024b226 <+1177>: mov %eax,0x4(%esp)
0x4024b22a <+1181>: mov 0x18(%esp),%eax
0x4024b22e <+1185>: mov %eax,(%esp)
0x4024b231 <+1188>: call 0x4021cbae <LDAge>
0x4024b236 <+1193>: jmp 0x4024b357 <TraceStart+1482>
0x4024b23b <+1198>: mov 0x18(%esp),%eax
0x4024b23f <+1202>: lea 0x4e0(%eax),%edi
0x4024b245 <+1208>: lea 0x500(%eax),%ecx
0x4024b24b <+1214>: mov %ecx,0x28(%esp)
0x4024b24f <+1218>: mov %eax,%esi
0x4024b251 <+1220>: mov %edi,0x10(%esp)
0x4024b255 <+1224>: mov 0x10(%esp),%eax
0x4024b259 <+1228>: mov %eax,%edi
0x4024b25b <+1230>: mov (%eax),%edx
0x4024b25d <+1232>: mov (%edx),%ebx
0x4024b25f <+1234>: cmp %eax,%edx
0x4024b261 <+1236>: je 0x4024b29d <TraceStart+1296>
0x4024b263 <+1238>: movzbl -0x3(%edx),%eax
0x4024b267 <+1242>: and $0x1,%eax
0x4024b26a <+1245>: test %eax,0x41c(%esi)
0x4024b270 <+1251>: jne 0x4024b291 <TraceStart+1284>
0x4024b272 <+1253>: mov 0x4(%ebp),%ecx
0x4024b275 <+1256>: bt %ecx,%eax
0x4024b278 <+1259>: jae 0x4024b291 <TraceStart+1284>
0x4024b27a <+1261>: movl $0x1,0x8(%esp)
0x4024b282 <+1269>: sub $0x1c,%edx
0x4024b285 <+1272>: mov %edx,0x4(%esp)
0x4024b289 <+1276>: mov %esi,(%esp)
0x4024b28c <+1279>: call 0x4024acdb <ShieldRaise>
0x4024b291 <+1284>: mov (%ebx),%eax
0x4024b293 <+1286>: cmp %ebx,%edi
0x4024b295 <+1288>: je 0x4024b29d <TraceStart+1296>
0x4024b297 <+1290>: mov %ebx,%edx
0x4024b299 <+1292>: mov %eax,%ebx
0x4024b29b <+1294>: jmp 0x4024b263 <TraceStart+1238>
0x4024b29d <+1296>: addl $0x8,0x10(%esp)
0x4024b2a2 <+1301>: mov 0x28(%esp),%eax
0x4024b2a6 <+1305>: cmp %eax,0x10(%esp)
0x4024b2aa <+1309>: jne 0x4024b255 <TraceStart+1224>
0x4024b2ac <+1311>: movl $0x3,0x18(%ebp)
0x4024b2b3 <+1318>: mov 0x4(%ebp),%ecx
0x4024b2b6 <+1321>: mov $0x1,%eax
0x4024b2bb <+1326>: shl %cl,%eax
0x4024b2bd <+1328>: mov 0x18(%esp),%ecx
0x4024b2c1 <+1332>: or %eax,0x41c(%ecx)
0x4024b2c7 <+1338>: mov 0x403011a8,%eax
0x4024b2cc <+1343>: sub $0x402fc1a0,%eax
0x4024b2d1 <+1348>: cmp $0x13,%eax
0x4024b2d4 <+1351>: ja 0x4024b30d <TraceStart+1408>
0x4024b2d6 <+1353>: movl $0x2,(%esp)
0x4024b2dd <+1360>: call 0x4022d788 <EventFlush>
0x4024b2e2 <+1365>: mov 0x403011a8,%eax
0x4024b2e7 <+1370>: sub $0x402fc1a0,%eax
0x4024b2ec <+1375>: cmp $0x13,%eax
0x4024b2ef <+1378>: ja 0x4024b30d <TraceStart+1408>
0x4024b2f1 <+1380>: movl $0x4025ef1c,0x8(%esp)
0x4024b2f9 <+1388>: movl $0x273,0x4(%esp)
0x4024b301 <+1396>: movl $0x40259c80,(%esp)
0x4024b308 <+1403>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b30d <+1408>: mov 0x403011a8,%ecx
0x4024b313 <+1414>: movw $0x21,-0x14(%ecx)
0x4024b319 <+1420>: movw $0x14,-0x12(%ecx)
0x4024b31f <+1426>: rdtsc
0x4024b321 <+1428>: mov %eax,-0x10(%ecx)
0x4024b324 <+1431>: mov %edx,-0xc(%ecx)
0x4024b327 <+1434>: mov %ebp,-0x8(%ecx)
0x4024b32a <+1437>: mov 0x18(%esp),%eax
0x4024b32e <+1441>: mov %eax,-0x4(%ecx)
0x4024b331 <+1444>: subl $0x14,0x403011a8
0x4024b338 <+1451>: mov %eax,(%esp)
0x4024b33b <+1454>: call 0x4021c8aa <ShieldResume>
0x4024b340 <+1459>: mov $0x0,%eax
0x4024b345 <+1464>: jmp 0x4024b3ad <TraceStart+1568>
0x4024b347 <+1466>: mov 0x18(%esp),%eax
0x4024b34b <+1470>: mov %eax,(%esp)
0x4024b34e <+1473>: call 0x4021c8aa <ShieldResume>
0x4024b353 <+1478>: mov %ebx,%eax
0x4024b355 <+1480>: jmp 0x4024b3ad <TraceStart+1568>
0x4024b357 <+1482>: movl $0x0,0x3c(%esp)
0x4024b35f <+1490>: lea 0x34(%esp),%eax
0x4024b363 <+1494>: mov %eax,0x8(%esp)
0x4024b367 <+1498>: movl $0x402471ad,0x4(%esp)
0x4024b36f <+1506>: mov 0x18(%esp),%edi
0x4024b373 <+1510>: mov %edi,(%esp)
0x4024b376 <+1513>: call 0x4020a32c <RootsIterate>
0x4024b37b <+1518>: mov %eax,%ebx
0x4024b37d <+1520>: test %eax,%eax
0x4024b37f <+1522>: jne 0x4024b347 <TraceStart+1466>
0x4024b381 <+1524>: movl $0x1,0x3c(%esp)
0x4024b389 <+1532>: lea 0x34(%esp),%eax
0x4024b38d <+1536>: mov %eax,0x8(%esp)
0x4024b391 <+1540>: movl $0x402471ad,0x4(%esp)
0x4024b399 <+1548>: mov %edi,(%esp)
0x4024b39c <+1551>: call 0x4020a32c <RootsIterate>
0x4024b3a1 <+1556>: mov %eax,%ebx
0x4024b3a3 <+1558>: test %eax,%eax
0x4024b3a5 <+1560>: je 0x4024b23b <TraceStart+1198>
0x4024b3ab <+1566>: jmp 0x4024b347 <TraceStart+1466>
0x4024b3ad <+1568>: add $0x4c,%esp
0x4024b3b0 <+1571>: pop %ebx
0x4024b3b1 <+1572>: pop %esi
0x4024b3b2 <+1573>: pop %edi
0x4024b3b3 <+1574>: pop %ebp
0x4024b3b4 <+1575>: ret
End of assembler dump.
(gdb) info float
R7: Zero 0x00000000000000000000 +0
R6: Zero 0x00000000000000000000 +0
R5: Zero 0x00000000000000000000 +0
R4: Zero 0x00000000000000000000 +0
R3: Zero 0x00000000000000000000 +0
R2: Zero 0x00000000000000000000 +0
R1: Valid 0x401683d14c0000000000 +8638796
=>R0: Valid 0x3fff8000000000000000 +1
Status Word: 0x83e1 IE PE ES SF C0 C1
TOP: 0
Control Word: 0x0362 DM PM
PC: Extended Precision (64-bits)
RC: Round to nearest
Tag Word: 0x5550
Instruction Pointer: 0x73:0x4024af8e
Operand Pointer: 0x7b:0xbfffd834
Opcode: 0xdd44
(gdb) |
Okay, you asked for it. I have a new theory: Perhaps with the HARP run-time, we aren't resetting the FPU status correctly after handling an error and that's causing something in MPS using floating point to think it had an error. This is pre- 0x4024af92 in TraceStart ()
from /home/bruce/Development/opendylan/Bootstrap.3/bin/../lib/libdylan.so
(gdb) disassemble
Dump of assembler code for function TraceStart:
0x4024ad8d <+0>: push %ebp
0x4024ad8e <+1>: push %edi
0x4024ad8f <+2>: push %esi
0x4024ad90 <+3>: push %ebx
0x4024ad91 <+4>: sub $0x4c,%esp
0x4024ad94 <+7>: mov 0x60(%esp),%ebp
0x4024ad98 <+11>: fldl 0x64(%esp)
0x4024ad9c <+15>: fstpl 0x10(%esp)
0x4024ada0 <+19>: fldl 0x6c(%esp)
0x4024ada4 <+23>: fstpl 0x18(%esp)
0x4024ada8 <+27>: mov %ebp,(%esp)
0x4024adab <+30>: call 0x40215afa <TraceCheck>
0x4024adb0 <+35>: test %eax,%eax
0x4024adb2 <+37>: jne 0x4024add0 <TraceStart+67>
0x4024adb4 <+39>: movl $0x40262b80,0x8(%esp)
0x4024adbc <+47>: movl $0x658,0x4(%esp)
0x4024adc4 <+55>: movl $0x40259c80,(%esp)
0x4024adcb <+62>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024add0 <+67>: cmpl $0x1,0x18(%ebp)
0x4024add4 <+71>: je 0x4024adf2 <TraceStart+101>
0x4024add6 <+73>: movl $0x40264d37,0x8(%esp)
0x4024adde <+81>: movl $0x659,0x4(%esp)
0x4024ade6 <+89>: movl $0x40259c80,(%esp)
0x4024aded <+96>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024adf2 <+101>: fldz
0x4024adf4 <+103>: fldl 0x10(%esp)
0x4024adf8 <+107>: fucomip %st(1),%st
0x4024adfa <+109>: fstp %st(0)
0x4024adfc <+111>: jae 0x4024ae1a <TraceStart+141>
0x4024adfe <+113>: movl $0x402661db,0x8(%esp)
0x4024ae06 <+121>: movl $0x65a,0x4(%esp)
0x4024ae0e <+129>: movl $0x40259c80,(%esp)
0x4024ae15 <+136>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae1a <+141>: fld1
0x4024ae1c <+143>: fldl 0x10(%esp)
0x4024ae20 <+147>: fxch %st(1)
0x4024ae22 <+149>: fucomip %st(1),%st
0x4024ae24 <+151>: fstp %st(0)
0x4024ae26 <+153>: jae 0x4024ae44 <TraceStart+183>
0x4024ae28 <+155>: movl $0x402661ec,0x8(%esp)
0x4024ae30 <+163>: movl $0x65b,0x4(%esp)
0x4024ae38 <+171>: movl $0x40259c80,(%esp)
0x4024ae3f <+178>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae44 <+183>: fldz
0x4024ae46 <+185>: fldl 0x18(%esp)
0x4024ae4a <+189>: fucomip %st(1),%st
0x4024ae4c <+191>: fstp %st(0)
0x4024ae4e <+193>: jae 0x4024ae6c <TraceStart+223>
0x4024ae50 <+195>: movl $0x402661fd,0x8(%esp)
0x4024ae58 <+203>: movl $0x65c,0x4(%esp)
0x4024ae60 <+211>: movl $0x40259c80,(%esp)
0x4024ae67 <+218>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024ae6c <+223>: mov 0x8(%ebp),%ebx
0x4024ae6f <+226>: mov %ebx,0x4(%esp)
0x4024ae73 <+230>: lea 0x30(%esp),%eax
0x4024ae77 <+234>: mov %eax,(%esp)
0x4024ae7a <+237>: call 0x4021c5ef <SegFirst>
0x4024ae7f <+242>: test %eax,%eax
0x4024ae81 <+244>: je 0x4024af44 <TraceStart+439>
0x4024ae87 <+250>: mov 0x30(%esp),%eax
0x4024ae8b <+254>: mov 0x14(%eax),%edi
0x4024ae8e <+257>: mov 0x8(%eax),%edx
0x4024ae91 <+260>: mov 0x8(%edx),%esi
0x4024ae94 <+263>: movzbl 0x19(%eax),%edx
0x4024ae98 <+267>: and $0x1,%edx
0x4024ae9b <+270>: mov 0x4(%ebp),%eax
0x4024ae9e <+273>: bt %eax,%edx
0x4024aea1 <+276>: jae 0x4024aebf <TraceStart+306>
0x4024aea3 <+278>: movl $0x4025c4e4,0x8(%esp)
0x4024aeab <+286>: movl $0x66b,0x4(%esp)
0x4024aeb3 <+294>: movl $0x40259c80,(%esp)
0x4024aeba <+301>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024aebf <+306>: mov 0x30(%esp),%eax
0x4024aec3 <+310>: testb $0x78,0x19(%eax)
0x4024aec7 <+314>: je 0x4024af24 <TraceStart+407>
0x4024aec9 <+316>: sub %esi,%edi
0x4024aecb <+318>: mov 0x10(%ebp),%edx
0x4024aece <+321>: test %edx,0x24(%eax)
0x4024aed1 <+324>: je 0x4024aefe <TraceStart+369>
0x4024aed3 <+326>: mov %eax,0x8(%esp)
0x4024aed7 <+330>: mov %ebp,0x4(%esp)
0x4024aedb <+334>: mov 0x8(%eax),%eax
0x4024aede <+337>: mov (%eax),%eax
0x4024aee0 <+339>: mov %eax,(%esp)
0x4024aee3 <+342>: call 0x40216cef <PoolGrey>
0x4024aee8 <+347>: mov 0x30(%esp),%eax
0x4024aeec <+351>: movzbl 0x19(%eax),%edx
0x4024aef0 <+355>: and $0x1,%edx
0x4024aef3 <+358>: mov 0x4(%ebp),%eax
0x4024aef6 <+361>: bt %eax,%edx
0x4024aef9 <+364>: jae 0x4024aefe <TraceStart+369>
0x4024aefb <+366>: add %edi,0x3c(%ebp)
0x4024aefe <+369>: mov 0x30(%esp),%eax
0x4024af02 <+373>: mov 0x8(%eax),%edx
0x4024af05 <+376>: mov (%edx),%edx
0x4024af07 <+378>: mov 0x8(%edx),%edx
0x4024af0a <+381>: testb $0x2,0x1c(%edx)
0x4024af0e <+385>: je 0x4024af24 <TraceStart+407>
0x4024af10 <+387>: movzbl 0x19(%eax),%edx
0x4024af14 <+391>: shr %dl
0x4024af16 <+393>: and $0x1,%edx
0x4024af19 <+396>: mov 0x4(%ebp),%eax
0x4024af1c <+399>: bt %eax,%edx
0x4024af1f <+402>: jb 0x4024af24 <TraceStart+407>
0x4024af21 <+404>: add %edi,0x38(%ebp)
0x4024af24 <+407>: mov 0x30(%esp),%eax
0x4024af28 <+411>: mov %eax,0x8(%esp)
0x4024af2c <+415>: mov %ebx,0x4(%esp)
0x4024af30 <+419>: lea 0x30(%esp),%eax
0x4024af34 <+423>: mov %eax,(%esp)
0x4024af37 <+426>: call 0x4021c5c0 <SegNext>
0x4024af3c <+431>: test %eax,%eax
0x4024af3e <+433>: jne 0x4024ae87 <TraceStart+250>
0x4024af44 <+439>: mov %ebp,0x8(%esp)
0x4024af48 <+443>: movl $0x4021668b,0x4(%esp)
0x4024af50 <+451>: mov %ebx,(%esp)
0x4024af53 <+454>: call 0x4020a32c <RootsIterate>
0x4024af58 <+459>: test %eax,%eax
0x4024af5a <+461>: je 0x4024af78 <TraceStart+491>
0x4024af5c <+463>: movl $0x40266315,0x8(%esp)
0x4024af64 <+471>: movl $0x689,0x4(%esp)
0x4024af6c <+479>: movl $0x40259c80,(%esp)
0x4024af73 <+486>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024af78 <+491>: mov 0x34(%ebp),%eax
0x4024af7b <+494>: mov $0x0,%edx
0x4024af80 <+499>: mov %eax,0x20(%esp)
0x4024af84 <+503>: mov %edx,0x24(%esp)
0x4024af88 <+507>: fildll 0x20(%esp)
0x4024af8c <+511>: fld1
0x4024af8e <+513>: fldl 0x10(%esp)
=> 0x4024af92 <+517>: fsubr %st(1),%st
0x4024af94 <+519>: fmulp %st,%st(2)
0x4024af96 <+521>: fxch %st(1)
0x4024af98 <+523>: fnstcw 0x2e(%esp)
0x4024af9c <+527>: movzwl 0x2e(%esp),%eax
0x4024afa1 <+532>: mov $0xc,%ah
0x4024afa3 <+534>: mov %ax,0x2c(%esp)
0x4024afa8 <+539>: fldcw 0x2c(%esp)
0x4024afac <+543>: fistpll 0x20(%esp)
0x4024afb0 <+547>: fldcw 0x2e(%esp)
0x4024afb4 <+551>: mov 0x20(%esp),%eax
0x4024afb8 <+555>: fldl 0x18(%esp)
0x4024afbc <+559>: fmuls 0x402678a8
0x4024afc2 <+565>: fxch %st(1)
0x4024afc4 <+567>: fucomip %st(1),%st
0x4024afc6 <+569>: ja 0x4024afd8 <TraceStart+587>
0x4024afc8 <+571>: fldl 0x402678e8
0x4024afce <+577>: fxch %st(1)
0x4024afd0 <+579>: fucomi %st(1),%st
0x4024afd2 <+581>: fcmovnb %st(1),%st
0x4024afd4 <+583>: fstp %st(1)
0x4024afd6 <+585>: jmp 0x4024afdc <TraceStart+591>
0x4024afd8 <+587>: fstp %st(0)
0x4024afda <+589>: fld1
0x4024afdc <+591>: add 0x3c(%ebp),%eax
0x4024afdf <+594>: fldcw 0x2c(%esp)
0x4024afe3 <+598>: fistpll 0x20(%esp)
0x4024afe7 <+602>: fldcw 0x2e(%esp)
0x4024afeb <+606>: mov 0x20(%esp),%ecx
0x4024afef <+610>: mov $0x0,%edx
0x4024aff4 <+615>: div %ecx
0x4024aff6 <+617>: add $0x1,%eax
0x4024aff9 <+620>: mov %eax,0x40(%ebp)
0x4024affc <+623>: mov 0x403011a8,%eax
0x4024b001 <+628>: sub $0x402fc1a0,%eax
0x4024b006 <+633>: cmp $0x33,%eax
0x4024b009 <+636>: ja 0x4024b042 <TraceStart+693>
0x4024b00b <+638>: movl $0x2,(%esp)
0x4024b012 <+645>: call 0x4022d788 <EventFlush>
0x4024b017 <+650>: mov 0x403011a8,%eax
0x4024b01c <+655>: sub $0x402fc1a0,%eax
0x4024b021 <+660>: cmp $0x33,%eax
0x4024b024 <+663>: ja 0x4024b042 <TraceStart+693>
0x4024b026 <+665>: movl $0x4025ee34,0x8(%esp)
0x4024b02e <+673>: movl $0x6a2,0x4(%esp)
0x4024b036 <+681>: movl $0x40259c80,(%esp)
0x4024b03d <+688>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b042 <+693>: mov 0x403011a8,%ecx
0x4024b048 <+699>: movw $0x1c,-0x34(%ecx)
0x4024b04e <+705>: movw $0x34,-0x32(%ecx)
0x4024b054 <+711>: rdtsc
0x4024b056 <+713>: mov %eax,-0x30(%ecx)
0x4024b059 <+716>: mov %edx,-0x2c(%ecx)
0x4024b05c <+719>: mov %ebp,-0x28(%ecx)
0x4024b05f <+722>: fldl 0x10(%esp)
0x4024b063 <+726>: fstpl -0x24(%ecx)
0x4024b066 <+729>: fldl 0x18(%esp)
0x4024b06a <+733>: fstpl -0x1c(%ecx)
0x4024b06d <+736>: mov 0x34(%ebp),%eax
0x4024b070 <+739>: mov %eax,-0x14(%ecx)
0x4024b073 <+742>: mov 0x38(%ebp),%eax
0x4024b076 <+745>: mov %eax,-0x10(%ecx)
0x4024b079 <+748>: mov 0x3c(%ebp),%eax
0x4024b07c <+751>: mov %eax,-0xc(%ecx)
0x4024b07f <+754>: mov 0x10(%ebp),%eax
0x4024b082 <+757>: mov %eax,-0x8(%ecx)
0x4024b085 <+760>: mov 0x40(%ebp),%eax
0x4024b088 <+763>: mov %eax,-0x4(%ecx)
0x4024b08b <+766>: subl $0x34,0x403011a8
0x4024b092 <+773>: movl $0x2,0x18(%ebp)
0x4024b099 <+780>: mov %ebp,(%esp)
0x4024b09c <+783>: call 0x4022c6fb <TracePostStartMessage>
0x4024b0a1 <+788>: mov %ebp,(%esp)
0x4024b0a4 <+791>: call 0x40215afa <TraceCheck>
0x4024b0a9 <+796>: test %eax,%eax
0x4024b0ab <+798>: jne 0x4024b0c9 <TraceStart+828>
0x4024b0ad <+800>: movl $0x40262b80,0x8(%esp)
0x4024b0b5 <+808>: movl $0x233,0x4(%esp)
0x4024b0bd <+816>: movl $0x40259c80,(%esp)
0x4024b0c4 <+823>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b0c9 <+828>: mov 0x4(%ebp),%ecx
0x4024b0cc <+831>: mov $0x1,%eax
0x4024b0d1 <+836>: shl %cl,%eax
0x4024b0d3 <+838>: mov %eax,0x34(%esp)
0x4024b0d7 <+842>: mov 0x8(%ebp),%eax
0x4024b0da <+845>: mov %eax,0x18(%esp)
0x4024b0de <+849>: mov %eax,0x38(%esp)
0x4024b0e2 <+853>: mov %eax,(%esp)
0x4024b0e5 <+856>: call 0x4024384c <ShieldSuspend>
0x4024b0ea <+861>: cmpl $0x2,0x18(%ebp)
0x4024b0ee <+865>: je 0x4024b10c <TraceStart+895>
0x4024b0f0 <+867>: movl $0x4025bd24,0x8(%esp)
0x4024b0f8 <+875>: movl $0x23a,0x4(%esp)
0x4024b100 <+883>: movl $0x40259c80,(%esp)
0x4024b107 <+890>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b10c <+895>: mov 0x4(%ebp),%eax
0x4024b10f <+898>: mov 0x18(%esp),%ecx
0x4024b113 <+902>: mov 0x41c(%ecx),%edx
0x4024b119 <+908>: bt %eax,%edx
0x4024b11c <+911>: jae 0x4024b13a <TraceStart+941>
0x4024b11e <+913>: movl $0x4025ee8c,0x8(%esp)
0x4024b126 <+921>: movl $0x23b,0x4(%esp)
0x4024b12e <+929>: movl $0x40259c80,(%esp)
0x4024b135 <+936>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b13a <+941>: mov 0x403011a8,%eax
0x4024b13f <+946>: sub $0x402fc1a0,%eax
0x4024b144 <+951>: cmp $0x13,%eax
0x4024b147 <+954>: ja 0x4024b180 <TraceStart+1011>
0x4024b149 <+956>: movl $0x2,(%esp)
0x4024b150 <+963>: call 0x4022d788 <EventFlush>
0x4024b155 <+968>: mov 0x403011a8,%eax
0x4024b15a <+973>: sub $0x402fc1a0,%eax
0x4024b15f <+978>: cmp $0x13,%eax
0x4024b162 <+981>: ja 0x4024b180 <TraceStart+1011>
0x4024b164 <+983>: movl $0x4025eebc,0x8(%esp)
0x4024b16c <+991>: movl $0x23d,0x4(%esp)
0x4024b174 <+999>: movl $0x40259c80,(%esp)
0x4024b17b <+1006>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b180 <+1011>: mov 0x403011a8,%ecx
0x4024b186 <+1017>: movw $0x20,-0x14(%ecx)
0x4024b18c <+1023>: movw $0x14,-0x12(%ecx)
0x4024b192 <+1029>: rdtsc
0x4024b194 <+1031>: mov %eax,-0x10(%ecx)
0x4024b197 <+1034>: mov %edx,-0xc(%ecx)
0x4024b19a <+1037>: mov %ebp,-0x8(%ecx)
0x4024b19d <+1040>: mov 0x18(%esp),%eax
0x4024b1a1 <+1044>: mov %eax,-0x4(%ecx)
0x4024b1a4 <+1047>: subl $0x14,0x403011a8
0x4024b1ab <+1054>: mov 0x50(%eax),%esi
0x4024b1ae <+1057>: mov (%esi),%edi
0x4024b1b0 <+1059>: add $0x50,%eax
0x4024b1b3 <+1062>: cmp %eax,%esi
0x4024b1b5 <+1064>: je 0x4024b21b <TraceStart+1166>
0x4024b1b7 <+1066>: mov %ebp,0x60(%esp)
0x4024b1bb <+1070>: mov %eax,%ebp
0x4024b1bd <+1072>: lea -0x10(%esi),%eax
0x4024b1c0 <+1075>: mov %eax,(%esp)
0x4024b1c3 <+1078>: call 0x40214d34 <PoolCheck>
0x4024b1c8 <+1083>: test %eax,%eax
0x4024b1ca <+1085>: jne 0x4024b1e8 <TraceStart+1115>
0x4024b1cc <+1087>: movl $0x40262a05,0x8(%esp)
0x4024b1d4 <+1095>: movl $0x1c0,0x4(%esp)
0x4024b1dc <+1103>: movl $0x40259c80,(%esp)
0x4024b1e3 <+1110>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b1e8 <+1115>: mov 0x8(%esi),%eax
0x4024b1eb <+1118>: mov (%eax),%ebx
0x4024b1ed <+1120>: add $0x8,%esi
0x4024b1f0 <+1123>: cmp %esi,%eax
0x4024b1f2 <+1125>: je 0x4024b20b <TraceStart+1150>
0x4024b1f4 <+1127>: sub $0x14,%eax
0x4024b1f7 <+1130>: mov %eax,(%esp)
0x4024b1fa <+1133>: call 0x4021916f <BufferFlip>
0x4024b1ff <+1138>: mov (%ebx),%edx
0x4024b201 <+1140>: cmp %ebx,%esi
0x4024b203 <+1142>: je 0x4024b20b <TraceStart+1150>
0x4024b205 <+1144>: mov %ebx,%eax
0x4024b207 <+1146>: mov %edx,%ebx
0x4024b209 <+1148>: jmp 0x4024b1f4 <TraceStart+1127>
0x4024b20b <+1150>: mov (%edi),%eax
0x4024b20d <+1152>: cmp %edi,%ebp
0x4024b20f <+1154>: je 0x4024b217 <TraceStart+1162>
0x4024b211 <+1156>: mov %edi,%esi
0x4024b213 <+1158>: mov %eax,%edi
0x4024b215 <+1160>: jmp 0x4024b1bd <TraceStart+1072>
0x4024b217 <+1162>: mov 0x60(%esp),%ebp
0x4024b21b <+1166>: mov 0x14(%ebp),%eax
0x4024b21e <+1169>: test %eax,%eax
0x4024b220 <+1171>: je 0x4024b357 <TraceStart+1482>
0x4024b226 <+1177>: mov %eax,0x4(%esp)
0x4024b22a <+1181>: mov 0x18(%esp),%eax
0x4024b22e <+1185>: mov %eax,(%esp)
0x4024b231 <+1188>: call 0x4021cbae <LDAge>
0x4024b236 <+1193>: jmp 0x4024b357 <TraceStart+1482>
0x4024b23b <+1198>: mov 0x18(%esp),%eax
0x4024b23f <+1202>: lea 0x4e0(%eax),%edi
0x4024b245 <+1208>: lea 0x500(%eax),%ecx
0x4024b24b <+1214>: mov %ecx,0x28(%esp)
0x4024b24f <+1218>: mov %eax,%esi
0x4024b251 <+1220>: mov %edi,0x10(%esp)
0x4024b255 <+1224>: mov 0x10(%esp),%eax
0x4024b259 <+1228>: mov %eax,%edi
0x4024b25b <+1230>: mov (%eax),%edx
0x4024b25d <+1232>: mov (%edx),%ebx
0x4024b25f <+1234>: cmp %eax,%edx
0x4024b261 <+1236>: je 0x4024b29d <TraceStart+1296>
0x4024b263 <+1238>: movzbl -0x3(%edx),%eax
0x4024b267 <+1242>: and $0x1,%eax
0x4024b26a <+1245>: test %eax,0x41c(%esi)
0x4024b270 <+1251>: jne 0x4024b291 <TraceStart+1284>
0x4024b272 <+1253>: mov 0x4(%ebp),%ecx
0x4024b275 <+1256>: bt %ecx,%eax
0x4024b278 <+1259>: jae 0x4024b291 <TraceStart+1284>
0x4024b27a <+1261>: movl $0x1,0x8(%esp)
0x4024b282 <+1269>: sub $0x1c,%edx
0x4024b285 <+1272>: mov %edx,0x4(%esp)
0x4024b289 <+1276>: mov %esi,(%esp)
0x4024b28c <+1279>: call 0x4024acdb <ShieldRaise>
0x4024b291 <+1284>: mov (%ebx),%eax
0x4024b293 <+1286>: cmp %ebx,%edi
0x4024b295 <+1288>: je 0x4024b29d <TraceStart+1296>
0x4024b297 <+1290>: mov %ebx,%edx
0x4024b299 <+1292>: mov %eax,%ebx
0x4024b29b <+1294>: jmp 0x4024b263 <TraceStart+1238>
0x4024b29d <+1296>: addl $0x8,0x10(%esp)
0x4024b2a2 <+1301>: mov 0x28(%esp),%eax
0x4024b2a6 <+1305>: cmp %eax,0x10(%esp)
0x4024b2aa <+1309>: jne 0x4024b255 <TraceStart+1224>
0x4024b2ac <+1311>: movl $0x3,0x18(%ebp)
0x4024b2b3 <+1318>: mov 0x4(%ebp),%ecx
0x4024b2b6 <+1321>: mov $0x1,%eax
0x4024b2bb <+1326>: shl %cl,%eax
0x4024b2bd <+1328>: mov 0x18(%esp),%ecx
0x4024b2c1 <+1332>: or %eax,0x41c(%ecx)
0x4024b2c7 <+1338>: mov 0x403011a8,%eax
0x4024b2cc <+1343>: sub $0x402fc1a0,%eax
0x4024b2d1 <+1348>: cmp $0x13,%eax
0x4024b2d4 <+1351>: ja 0x4024b30d <TraceStart+1408>
0x4024b2d6 <+1353>: movl $0x2,(%esp)
0x4024b2dd <+1360>: call 0x4022d788 <EventFlush>
0x4024b2e2 <+1365>: mov 0x403011a8,%eax
0x4024b2e7 <+1370>: sub $0x402fc1a0,%eax
0x4024b2ec <+1375>: cmp $0x13,%eax
0x4024b2ef <+1378>: ja 0x4024b30d <TraceStart+1408>
0x4024b2f1 <+1380>: movl $0x4025ef1c,0x8(%esp)
0x4024b2f9 <+1388>: movl $0x273,0x4(%esp)
0x4024b301 <+1396>: movl $0x40259c80,(%esp)
0x4024b308 <+1403>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024b30d <+1408>: mov 0x403011a8,%ecx
0x4024b313 <+1414>: movw $0x21,-0x14(%ecx)
0x4024b319 <+1420>: movw $0x14,-0x12(%ecx)
0x4024b31f <+1426>: rdtsc
0x4024b321 <+1428>: mov %eax,-0x10(%ecx)
0x4024b324 <+1431>: mov %edx,-0xc(%ecx)
0x4024b327 <+1434>: mov %ebp,-0x8(%ecx)
0x4024b32a <+1437>: mov 0x18(%esp),%eax
0x4024b32e <+1441>: mov %eax,-0x4(%ecx)
0x4024b331 <+1444>: subl $0x14,0x403011a8
0x4024b338 <+1451>: mov %eax,(%esp)
0x4024b33b <+1454>: call 0x4021c8aa <ShieldResume>
0x4024b340 <+1459>: mov $0x0,%eax
0x4024b345 <+1464>: jmp 0x4024b3ad <TraceStart+1568>
0x4024b347 <+1466>: mov 0x18(%esp),%eax
0x4024b34b <+1470>: mov %eax,(%esp)
0x4024b34e <+1473>: call 0x4021c8aa <ShieldResume>
0x4024b353 <+1478>: mov %ebx,%eax
0x4024b355 <+1480>: jmp 0x4024b3ad <TraceStart+1568>
0x4024b357 <+1482>: movl $0x0,0x3c(%esp)
0x4024b35f <+1490>: lea 0x34(%esp),%eax
0x4024b363 <+1494>: mov %eax,0x8(%esp)
0x4024b367 <+1498>: movl $0x402471ad,0x4(%esp)
0x4024b36f <+1506>: mov 0x18(%esp),%edi
0x4024b373 <+1510>: mov %edi,(%esp)
0x4024b376 <+1513>: call 0x4020a32c <RootsIterate>
0x4024b37b <+1518>: mov %eax,%ebx
0x4024b37d <+1520>: test %eax,%eax
0x4024b37f <+1522>: jne 0x4024b347 <TraceStart+1466>
0x4024b381 <+1524>: movl $0x1,0x3c(%esp)
0x4024b389 <+1532>: lea 0x34(%esp),%eax
0x4024b38d <+1536>: mov %eax,0x8(%esp)
0x4024b391 <+1540>: movl $0x402471ad,0x4(%esp)
0x4024b399 <+1548>: mov %edi,(%esp)
0x4024b39c <+1551>: call 0x4020a32c <RootsIterate>
0x4024b3a1 <+1556>: mov %eax,%ebx
0x4024b3a3 <+1558>: test %eax,%eax
0x4024b3a5 <+1560>: je 0x4024b23b <TraceStart+1198>
0x4024b3ab <+1566>: jmp 0x4024b347 <TraceStart+1466>
0x4024b3ad <+1568>: add $0x4c,%esp
0x4024b3b0 <+1571>: pop %ebx
0x4024b3b1 <+1572>: pop %esi
0x4024b3b2 <+1573>: pop %edi
0x4024b3b3 <+1574>: pop %ebp
0x4024b3b4 <+1575>: ret
End of assembler dump.
(gdb) info float
R7: Zero 0x00000000000000000000 +0
R6: Zero 0x00000000000000000000 +0
R5: Zero 0x00000000000000000000 +0
R4: Zero 0x00000000000000000000 +0
R3: Zero 0x00000000000000000000 +0
R2: Zero 0x00000000000000000000 +0
R1: Valid 0x401683d14c0000000000 +8638796
=>R0: Valid 0x3fff8000000000000000 +1
Status Word: 0x83e1 IE PE ES SF C0 C1
TOP: 0
Control Word: 0x0362 DM PM
PC: Extended Precision (64-bits)
RC: Round to nearest
Tag Word: 0x5550
Instruction Pointer: 0x73:0x4024af8e
Operand Pointer: 0x7b:0xbfffd834
Opcode: 0xdd44
(gdb) |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
When in doubt, disassemble (not dissemble!).
(gdb) disassemble
Dump of assembler code for function dylan_float_invalid_handler:
=> 0x402091ca <+0>: push %ebp
0x402091cb <+1>: mov %esp,%ebp
0x402091cd <+3>: pushf
0x402091ce <+4>: push %ebx
0x402091cf <+5>: push %esi
0x402091d0 <+6>: push %edi
0x402091d1 <+7>: fnclex
0x402091d3 <+9>: call 0x401ace33 <Kfloat_invalidVKeI>
0x402091d8 <+14>: lea -0x10(%ebp),%esp
0x402091db <+17>: pop %edi
0x402091dc <+18>: pop %esi
0x402091dd <+19>: pop %ebx
0x402091de <+20>: popf
0x402091df <+21>: pop %ebp
0x402091e0 <+22>: ret
End of assembler dump.And indeed, we can set a breakpoint on Kfloat_invalidVKeI and when we're in there, we can see that the FPU got reset:
(gdb) info float
R7: Zero 0x00000000000000000000 +0
R6: Zero 0x00000000000000000000 +0
R5: Zero 0x00000000000000000000 +0
R4: Zero 0x00000000000000000000 +0
=>R3: Zero 0x00000000000000000000 +0
R2: Empty 0x401887e0bc0000000000
R1: Empty 0x00000000000000000000
R0: Empty 0x00000000000000000000
Status Word: 0x1900 C0
TOP: 3
Control Word: 0x0362 DM PM
PC: Extended Precision (64-bits)
RC: Round to nearest
Tag Word: 0x557f
Instruction Pointer: 0x73:0x401ac80b
Operand Pointer: 0x7b:0xbfffd80c
Opcode: 0xd875
|
When in doubt, disassemble (not dissemble!). (gdb) disassemble
Dump of assembler code for function dylan_float_invalid_handler:
=> 0x402091ca <+0>: push %ebp
0x402091cb <+1>: mov %esp,%ebp
0x402091cd <+3>: pushf
0x402091ce <+4>: push %ebx
0x402091cf <+5>: push %esi
0x402091d0 <+6>: push %edi
0x402091d1 <+7>: fnclex
0x402091d3 <+9>: call 0x401ace33 <Kfloat_invalidVKeI>
0x402091d8 <+14>: lea -0x10(%ebp),%esp
0x402091db <+17>: pop %edi
0x402091dc <+18>: pop %esi
0x402091dd <+19>: pop %ebx
0x402091de <+20>: popf
0x402091df <+21>: pop %ebp
0x402091e0 <+22>: ret
End of assembler dump.And indeed, we can set a breakpoint on |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
I wonder if the difference between fclex and fnclex might be important here: http://x86.renejeschke.de/html/file_module_x86_id_85.html
|
I wonder if the difference between |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
So, let's look at the MPS code in more detail ... I wish I had debug info for it! (I'm not sure why it gets lost).
0x4024af53 <+454>: call 0x4020a32c <RootsIterate>
0x4024af58 <+459>: test %eax,%eax
0x4024af5a <+461>: je 0x4024af78 <TraceStart+491>
0x4024af5c <+463>: movl $0x40266315,0x8(%esp)
0x4024af64 <+471>: movl $0x689,0x4(%esp)
0x4024af6c <+479>: movl $0x40259c80,(%esp)
0x4024af73 <+486>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024af78 <+491>: mov 0x34(%ebp),%eax
0x4024af7b <+494>: mov $0x0,%edx
0x4024af80 <+499>: mov %eax,0x20(%esp)
0x4024af84 <+503>: mov %edx,0x24(%esp)
0x4024af88 <+507>: fildll 0x20(%esp)
0x4024af8c <+511>: fld1
0x4024af8e <+513>: fldl 0x10(%esp)
=> 0x4024af92 <+517>: fsubr %st(1),%st
0x4024af94 <+519>: fmulp %st,%st(2)
0x4024af96 <+521>: fxch %st(1)
0x4024af98 <+523>: fnstcw 0x2e(%esp)
0x4024af9c <+527>: movzwl 0x2e(%esp),%eax
0x4024afa1 <+532>: mov $0xc,%ah
0x4024afa3 <+534>: mov %ax,0x2c(%esp)
0x4024afa8 <+539>: fldcw 0x2c(%esp)
0x4024afac <+543>: fistpll 0x20(%esp)
0x4024afb0 <+547>: fldcw 0x2e(%esp)
0x4024afb4 <+551>: mov 0x20(%esp),%eax
0x4024afb8 <+555>: fldl 0x18(%esp)
0x4024afbc <+559>: fmuls 0x402678a8
0x4024afc2 <+565>: fxch %st(1)
0x4024afc4 <+567>: fucomip %st(1),%st
0x4024afc6 <+569>: ja 0x4024afd8 <TraceStart+587>So, we can see that this is something that happens after a call to RootsIterate and after a check ... and that it does a subtraction and then a multiplication and it is the fsubr %st(1),%st that failed.
If we look in the C code, we see this in TraceStart:
res = RootsIterate(ArenaGlobals(arena), rootGrey, (void *)trace);
AVER(res == ResOK);
STATISTIC_STAT(EVENT2(ArenaWriteFaults, arena, arena->writeBarrierHitCount));
/* Calculate the rate of scanning. */
{
Size sSurvivors = (Size)(trace->condemned * (1.0 - mortality));
double nPolls = finishingTime / ArenaPollALLOCTIME;So .. I think it failed on 1.0 - mortality and, if we look at the floating point stack, we see:
R1: Valid 0x401683d14c0000000000 +8638796
=>R0: Valid 0x3fff8000000000000000 +1
So, should fsubr %st(1),%st have failed here?
|
So, let's look at the MPS code in more detail ... I wish I had debug info for it! (I'm not sure why it gets lost). 0x4024af53 <+454>: call 0x4020a32c <RootsIterate>
0x4024af58 <+459>: test %eax,%eax
0x4024af5a <+461>: je 0x4024af78 <TraceStart+491>
0x4024af5c <+463>: movl $0x40266315,0x8(%esp)
0x4024af64 <+471>: movl $0x689,0x4(%esp)
0x4024af6c <+479>: movl $0x40259c80,(%esp)
0x4024af73 <+486>: call 0x4020aa19 <mps_lib_assert_fail>
0x4024af78 <+491>: mov 0x34(%ebp),%eax
0x4024af7b <+494>: mov $0x0,%edx
0x4024af80 <+499>: mov %eax,0x20(%esp)
0x4024af84 <+503>: mov %edx,0x24(%esp)
0x4024af88 <+507>: fildll 0x20(%esp)
0x4024af8c <+511>: fld1
0x4024af8e <+513>: fldl 0x10(%esp)
=> 0x4024af92 <+517>: fsubr %st(1),%st
0x4024af94 <+519>: fmulp %st,%st(2)
0x4024af96 <+521>: fxch %st(1)
0x4024af98 <+523>: fnstcw 0x2e(%esp)
0x4024af9c <+527>: movzwl 0x2e(%esp),%eax
0x4024afa1 <+532>: mov $0xc,%ah
0x4024afa3 <+534>: mov %ax,0x2c(%esp)
0x4024afa8 <+539>: fldcw 0x2c(%esp)
0x4024afac <+543>: fistpll 0x20(%esp)
0x4024afb0 <+547>: fldcw 0x2e(%esp)
0x4024afb4 <+551>: mov 0x20(%esp),%eax
0x4024afb8 <+555>: fldl 0x18(%esp)
0x4024afbc <+559>: fmuls 0x402678a8
0x4024afc2 <+565>: fxch %st(1)
0x4024afc4 <+567>: fucomip %st(1),%st
0x4024afc6 <+569>: ja 0x4024afd8 <TraceStart+587>So, we can see that this is something that happens after a call to If we look in the C code, we see this in res = RootsIterate(ArenaGlobals(arena), rootGrey, (void *)trace);
AVER(res == ResOK);
STATISTIC_STAT(EVENT2(ArenaWriteFaults, arena, arena->writeBarrierHitCount));
/* Calculate the rate of scanning. */
{
Size sSurvivors = (Size)(trace->condemned * (1.0 - mortality));
double nPolls = finishingTime / ArenaPollALLOCTIME;So .. I think it failed on So, should |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
Okay. If we look at an info float in the first FP exception, we see that the status is:
Status Word: 0xb9a1 IE PE ES C0
TOP: 7
When we get into the MPS failure, it is:
Status Word: 0x83e1 IE PE ES SF C0 C1
TOP: 0
Differences: SF is set, as well as C1.
From http://x86.renejeschke.de/html/file_module_x86_id_122.html, fsubr will set C1 when:
C1 Set to 0 if stack underflow occurred. Set if result was rounded up; cleared otherwise. C0, C2, C3 Undefined.
|
Okay. If we look at an When we get into the MPS failure, it is: Differences: From http://x86.renejeschke.de/html/file_module_x86_id_122.html,
|
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
So, each time we hit our SIGFPE handler, we're losing an FPU stack slot ... so this is on our side.
|
So, each time we hit our |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
We need to more aggressively reset FPU state. I'm pretty sure this issue will impact the FPU exception handling in both the HARP and LLVM back-ends / run-times.
I'll probably convert HARP over to using <fenv.h> like LLVM is ...
|
We need to more aggressively reset FPU state. I'm pretty sure this issue will impact the FPU exception handling in both the HARP and LLVM back-ends / run-times. I'll probably convert HARP over to using |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment
Hide comment
waywardmonkeys
Apr 16, 2015
Member
I've been digging into this more and have been talking to some people including stassats and pkhuong from SBCL. (Earlier, I talked with nsz from MUSL libc.)
What is happening here is that FE_INVALID, on x86 and against what IEEE specifies, is what gets reported when there is an x87 stack fault. We may be able to ignore this and let things carry on. That requires some investigation. If so, we'll want to be looking at bit 6 of the FPU status word and if that's set, it is a stack fault. stassats pasted something that indicates that while SF is always set for a stack fault, it might also have stuck around and show up set when we're handling a real invalid FPE.
Of course, if we're using the SSE instructions and ignoring the x87 FPU, all is well and this isn't a concern.
|
I've been digging into this more and have been talking to some people including stassats and pkhuong from SBCL. (Earlier, I talked with nsz from MUSL libc.) What is happening here is that Of course, if we're using the SSE instructions and ignoring the x87 FPU, all is well and this isn't a concern. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
waywardmonkeys commentedApr 16, 2015
After the addition of FE_INVALID support in #874, the HARP builds using MPS now fail
make checkas a result of crashing while compilingcommon-dylan-test-suite-app.This is actually interesting.
The assertion failure from MPS in the build logs looks like this:
I installed a clean Ubuntu 14.04 VM and did a build using MPS 1.114.0 and current master of Open Dylan (and gcc 4.8.2).
After the
makecompleted, I ranmake check, let it crash, and then ranmake check FDMAKE_OPTIONS=--gdbso that it would run the failing build process under gdb.Then, I set up signal handling in gdb correctly for MPS:
And then I ran and it broke into the debugger here:
We can see here that it was doing a division operator on a float (
KSVKdMM4I) and that's where it hit the floating point exception. It did this in the context of attempting to constant fold some function calls during the optimization process.I
conta couple of times to get through this set of exceptions and then:I'm not sure why we hit this code path here in the GC (MPS people?), but we did.
I
contfrom there and:At that point, the stack trace looks like:
So, we can see that (as is usual for a
lockli.c, res == 0assertion failure) we've re-entered MPS from the signal handler. We were inside ofTraceStart, hit the FP exception, and then in the signal handler, we re-entered MPS to allocate the error object which allocated and caused a re-entry into MPS.