SIGSEGV error.
lazarus / 13-General
----------------------------------------------------------------------
Elphi · 2005-12-29 16:38
Hello,
When I run an application an exception is raised:
External: SISSEGV
I haven't got a clue where I should be looking for. What could be the
cause of this?
----------------------------------------------------------------------
matthijs · 2005-12-29 18:50
Is this with any application you run or only with a certain
application?
If the latter, can you reproduce this in a small example? Going by the
information you've given now, we can only guess.
----------------------------------------------------------------------
Anonymous · 2005-12-30 10:18
It is my first Lazarus application. The error surfaces when I include
a unit in the uses clause of the main form. That unit uses a lot of
units from my Delphi library which are ported in a quick and dirty
manner.
The IDE/application has been problemstic from the start. See for
instance http://www.lazarus.freepascal.org/index.php?name=PNphpBB2&fil
e=viewtopic&t=1729
for example for another hurdle in this venture.
Hope somebody has some hints what might be wrong.
----------------------------------------------------------------------
Elphi · 2005-12-30 11:31
Solved?
In the initialization section of one of the library units the CPU type
is determined.
The header of the function used looked like this:
Code: [Select]function CPU: Integer; assembler;
asm
end;
This compiled under:
Code: [Select]{$mode objfpc}
{$asmmode intel}
Changed to:
Code: [Select]function CPU: Integer;
begin
asm
end;
end;
and I get no exception anymore.
----------------------------------------------------------------------
matthijs · 2005-12-30 11:49
That's something we could not guess. But you are correct, one needs
begin ... end around asm-blocks if you are using objfpc-mode.
----------------------------------------------------------------------
Anonymous · 2005-12-30 12:01
Should it compile? I don't think so.
----------------------------------------------------------------------
Vincent Snijders · 2005-12-30 12:06
It depends on the body. Maybe you make assumptions in it
that are not correct, if (begin..end) is missing, for example
assumptions about the stack frame.
----------------------------------------------------------------------
Elphi · 2006-01-02 09:13
Im my view it is a typical case of syntax error that the compiler
should handle.
Here is the complete orginal routine including body:
Code: [Select]function CPU: Integer; assembler;
asm
push bx // Save bx
// cpu086 if bits 12-15 of flags are set
pushf // Save flags
pop bx // Save flags in bx
mov ax,$0fff // Clear bits 12-15
and ax,bx // in flags
push ax // Store new flags value on stack
popf // Replace current flags value
pushf // Set new flags
pop ax // Store new flags in ax
and ax,$0f000 // If bits 12-15 are set then the
CPU
cmp ax,$0f000 // is (at least) a 8086/8088
jne @186 // Higher processor
xor ax,ax // A 8086/8088: set family bits
11-8 to 000b
// Here test for FPU?
jmp @Exit
//cpu186 How?
@186:
// cpu286 if bits 12-15 are always clear
@286: or bx,$0f000 // Try to set bits 12-15
push bx
popf
pushf
pop ax
pop bx // Restore bx as 16 bits tests
will finish
and ax,$0f000 // If bits 12-15 aren't cleared
jnz @386 // then it is a higher processor
mov ax,$200 // A 286: set family bits 11-8 to
010b
// Here test for FPU?
jmp @Exit
// cpu386 if ac bit (bit 18) of eflags, introduced on the
cpu486 can not be set
@386: pushfd
pop eax // Get original eflags
mov ecx,eax // Save original eflags
xor eax,$40000 // Flip AC bit in eflags
push eax // Save for eflags
popfd // Copy to eflags
pushfd // Push eflags
pop eax // Get new eflags value
xor eax,ecx // If ac bit can be toggled
jne @486 // then it is a higher processor
mov eax,$300 // A 386: set family bits 11-8 to
011b
// Here test for FPU?
jmp @Exit
// cpu486 id id bit (bit 21) of eflags, introduced on the
cpu586 can not be set
@486: pushfd // Push original eflags
pop eax // Get original eflags in eax
mov ecx,eax // Save original eflags in ecx
or eax,$200000 // Flip id bit in eflags
push eax // Save eax for eflags
popfd // Copy to eflags
pushfd // Push eflags
pop eax // Get new eflags value
xor eax,ecx // if id bit can be changed
jne @586 // then it is a higher processor
mov eax,$400 // then the CPU is a 486
// Here test for FPU?
jmp @Exit
// cpu586 and higher: cpuid instruction
@586: push ebx // Save ebx as it is affected by
cpuid
xor eax,eax // Prepare for getting the number
of cpuid functions
db $0f,$0a2 // Cpuid instruction. Get number
of cpuid function
pop ebx // Restore ebx
test eax,eax // If number > 0 then
jnz @AutoF // use cpuid to get features
mov eax,$500 // If number = 0 then no
functions available
// "Manually" setting features: -
FPU (bit
jmp @Exit
@AutoF: push ebx // Save ebx as it is affected by
cpuid
mov eax,1 // Prepare getting function 1
information
db $0f,$0a2 // Cpuid instruction. Get family
and features
pop ebx // Restore ebx
and eax,$000000f00 // Family
and edx,$0fffff0ff // Features flags
or eax,edx // Combine family and features
@Exit:
end; {CPU}
This case revealed other unexpected behavior:
Code: [Select]
function CPU: Integer; {$ifndef fpc}assembler;
{$else}
begin
{$endif}
asm
{$ifdef fpc}
end;
{$endif}
end;
doesn't work. One has to rewrite the complete routine in fpc syntax.
----------------------------------------------------------------------
Vincent Snijders · 2006-01-02 10:00
Quote from: "Elphi"One has to rewrite the complete routine in fpc
syntax.
That would be much work. Since it seems you are porting existing code,
did you try adding {$mode Delphi} to the top of you source file?
Vincent
----------------------------------------------------------------------
Elphi · 2006-01-02 11:12
No, mode objfpc. The application I'm writing is build from scratch.
However it uses my Delphi library units extensively.
The reasons why I'm using mode objfpc:
1. It was recommended somewhere in the docs to use mode Delphi only
for the q&d ports. I have no intention to my leave my code in q&d
state.
2. Considering the advice mentioned above, I expect mode Delphi to
have drawbacks on size and speed of executables.
3. The main reason to use fpc right now is that it can build 64 bits
executables.
Please correct me if I'm wrong.
In the mean time, I've seen this SIGSEGV error message more than once,
so I expect the compiler is not yet ready to relentlessly throw syntax
errors; mode delphi or mode objfpc shouldn't be a factor in this, I
think.
----------------------------------------------------------------------
Vincent Snijders · 2006-01-02 11:56
support of mode Delphi is mostly syntax related. You can get the same
size and speed as mode objfpc, although the compiler developers think
that objfpc is more readable and maintainable. (You may disagree).
Maybe you can discuss your issues on the fpc-pascal mailing list.
----------------------------------------------------------------------
Anonymous · 2006-01-02 12:15
So, basically what you are saying is that when I plan to use code
interchangeably between Delphi and Lazarus mode delphi is advised,
even when 64 bits/Linux is the main focus when using Lazarus?
P.s. Unfortunately I'm one of those who have troubles reaching the fpc
pages reading/posting on their forums.
----------------------------------------------------------------------
Vincent Snijders · 2006-01-02 12:27
Yes, if you are going to use the same source for both Delphi / win32
and Lazarus / linux / 64 bits, then I would recommand using mode
Delphi.
It should not have an effect on the code generation. The mode settings
is purely syntax and setting defaults, like {$H+} and {$asmmode
intel}, which are different in {mode objfpc}.
Too bad, you cannot reach www.freepascal.org, so you cannot submit bug
reports about syntax problems either.
----------------------------------------------------------------------
Anonymous · 2006-01-02 17:24
It's very annoying indeed when the fpc (community) pages are not
reacheable or very slow. Did I notice an improvement today?
I use Delphi mode now in all units, but I'm still flooded by messages
concerning missing fpc include files.