OK.
I use gmail and so I didn't see the problem.
My thought was to keep these JIT bugs in same thread, but you are right.
Jussi
On 6 July 2012 18:44, Benoît Minisini wrote:
> Le 06/07/2012 17:35, Jussi Lahtinen a écrit :
> > Just quick info before I leave for couple days...
> > ...
>
> Can
Le 06/07/2012 17:35, Jussi Lahtinen a écrit :
> Just quick info before I leave for couple days...
> ...
Can you start a new thread for new bugs ? It becomes hard to read the
current one now, it is too deep!
Regards,
--
Benoît Minisini
---
Next bug... it's crash and I don't have any clue what causes it.
I'm trying to isolate it, but meanwhile here is bt from gdb:
Program received signal SIGSEGV, Segmentation fault.
0x7fffe3b0e430 in llvm::BasicBlock::getContext() const ()
from /usr/lib/libLLVM-3.2svn.so
(gdb) bt
#0 0x7ff
Yes, it works now!
Jussi
On 27 June 2012 00:12, Emil Lenngren wrote:
> It seems like when LLVM reads boolean pointers, it assumes the seven most
> significant bits are always 0, which is not the case in Gambas. Does the
> latest revision work? I have no opportunity to even compile it and see
It seems like when LLVM reads boolean pointers, it assumes the seven most
significant bits are always 0, which is not the case in Gambas. Does the
latest revision work? I have no opportunity to even compile it and see if
it works at the moment :)
/Emil
2012/6/26 Jussi Lahtinen
> .../MagicBug$ G
.../MagicBug$ GB_JIT=info gbx3
gb.jit: using LLVM 3.2.
gb.jit: beginning compiling MainModule.MainModule:
-
I am in Italy now on holiday so I cannot debug the code, but could you
please run gbx3 with GB_JIT=info and post the llvm code?
/Emil
2012/6/26 Jussi Lahtinen
> OK, here is more. This is really really weird.
>
> I expect message "Everything works!", but I get *only* "Something broke!!".
> But i
OK, here is more. This is really really weird.
I expect message "Everything works!", but I get *only* "Something broke!!".
But if I add 'Return', under Message("Everything works!"), or I remove
'Fast' then everything works as expected.
See attachment.
Gambas 3 rev 4860 @ Xubuntu 12.04 64bit
Jus
Le 19/06/2012 20:58, Emil Lenngren a écrit :
> Hi. Kiitos for another bug report :)
>
> Hmm.. I forgot that classes can override extern methods as well. That's
> where the real problem is.
>
> /Emil
>
Mmm. Seems weird at first sight, but finally it could be useful.
--
Benoît Minisini
That bug was actually an interpreter bug, but it is now corrected in the
latest revision :)
/Emil
2012/6/20 Jussi Lahtinen
> Yeah, I'm also leaving to countryside for Midsummer parties... but I made
> quick test,
> and I found another bug (see attachment).
>
> Similar problem with argument coun
Yeah, I'm also leaving to countryside for Midsummer parties... but I made
quick test,
and I found another bug (see attachment).
Similar problem with argument counting.
Jussi
On 20 June 2012 19:14, Emil Lenngren wrote:
> Tomorrow I will travel away for a week so I cannot do anything then, b
Tomorrow I will travel away for a week so I cannot do anything then, but
just keep sending emails and I will reply when I come home :)
/Emil
2012/6/20 Jussi Lahtinen
> So far everything seems to work!
> I'll continue further testing later.
>
> Jussi
>
>
>
> On 20 June 2012 01:10, Emil Lenngren
So far everything seems to work!
I'll continue further testing later.
Jussi
On 20 June 2012 01:10, Emil Lenngren wrote:
> Is revision #4851 useful for you?
> You can now convert functions in the current running class to pointers.
>
> /Emil
>
> 2012/6/19 Emil Lenngren
>
> > It should work in
Is revision #4851 useful for you?
You can now convert functions in the current running class to pointers.
/Emil
2012/6/19 Emil Lenngren
> It should work in the latest revision :)
>
>
> 2012/6/19 Emil Lenngren
>
>> If you want a quick workaround, you can have non-fast method that creates
>> a c
It should work in the latest revision :)
2012/6/19 Emil Lenngren
> If you want a quick workaround, you can have non-fast method that creates
> a callback, and immediately returns it, like:
>
> Private Function GenerateCallback() As Pointer
> Return CallbackFunction
> End
>
> Fast Private Funct
If you want a quick workaround, you can have non-fast method that creates a
callback, and immediately returns it, like:
Private Function GenerateCallback() As Pointer
Return CallbackFunction
End
Fast Private Function CallbackFunction()
...
End
Extern extfunc(cb As Pointer) In "somelib"
Fast P
No problem, I'm glad to help making Gambas more robust, when ever I have
time to do so!
And testing things seems to be easy contribution for me...
My project has about 120 000 lines of code. And I try JIT with every module
and class of it.
So I think I will find more bugs... But in some point I mu
Hi. Kiitos for another bug report :)
Hmm.. I forgot that classes can override extern methods as well. That's
where the real problem is.
/Emil
2012/6/19 Jussi Lahtinen
> More bugs...
>
> Make shared library (just dummy example [or just use binary that is
> included with project source]);
>
> #i
More bugs...
Make shared library (just dummy example [or just use binary that is
included with project source]);
#include
int yEta(int x1, int y1, int x2, int y2)
{return (int) hypot(x1-x2,y1-y2);}
Then run the attached project.
With JIT, number of arguments is counted wrong.
Jussi
Yes, it works now!
Jussi
On 18 June 2012 22:34, Emil Lenngren wrote:
> Does it work in revision #4847?
>
> /Emil
>
> 2012/6/18 Jussi Lahtinen
>
> > Starting program: /usr/local/bin/gbx3
> > [Thread debugging using libthread_db enabled]
> > Using host libthread_db library
> "/lib/x86_64-linux
Does it work in revision #4847?
/Emil
2012/6/18 Jussi Lahtinen
> Starting program: /usr/local/bin/gbx3
> [Thread debugging using libthread_db enabled]
> Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> gbx3: ExecutionEngine.cpp:165: void
> llvm::ExecutionEngine::addG
Starting program: /usr/local/bin/gbx3
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
gbx3: ExecutionEngine.cpp:165: void
llvm::ExecutionEngine::addGlobalMapping(const llvm::GlobalValue*, void*):
Assertion `(CurVal == 0 || Add
Hi. Can you please provide more information? A gdb backtrace or something,
because I get no errors at all. Everything seems to work for me ...
/Emil
2012/6/18 Jussi Lahtinen
> OK, I found it. See attachment.
>
> Jussi
>
>
>
>
> On 17 June 2012 23:36, Jussi Lahtinen wrote:
>
> > OK, that is fix
OK, I found it. See attachment.
Jussi
On 17 June 2012 23:36, Jussi Lahtinen wrote:
> OK, that is fixed, and I cannot see any problems with compare method
> anymore.
>
> But with my big project, I got this with signal 6;
> gbx3: ExecutionEngine.cpp:165: void
> llvm::ExecutionEngine::addGlobal
OK, that is fixed, and I cannot see any problems with compare method
anymore.
But with my big project, I got this with signal 6;
gbx3: ExecutionEngine.cpp:165: void
llvm::ExecutionEngine::addGlobalMapping(const llvm::GlobalValue*, void*):
Assertion `(CurVal == 0 || Addr == 0) && "GlobalMapping alr
It should be fixed in revision #4843!
/Emil
2012/6/17 Jussi Lahtinen
> Yes, it really works now. And I found a lot of errors by me.
>
> But for some reason JIT raises compare method when it shouldn't,
> and when I tried to isolate the problem I got crash (signal 11).
> This seems not to be rel
Yes, it really works now. And I found a lot of errors by me.
But for some reason JIT raises compare method when it shouldn't,
and when I tried to isolate the problem I got crash (signal 11).
This seems not to be related to compare method, so just ignore name of the
attached project.
I will look i
Yes, it works now :)
2012/6/17 Benoît Minisini
> Le 17/06/2012 02:54, Jussi Lahtinen a écrit :
> > Yeah, terminal gives other position...
> > With my actual project, I got error message; wanted string, got
> > blaablaaClass instead.
> > And that error message comes from function (from other modu
Le 17/06/2012 02:54, Jussi Lahtinen a écrit :
> Yeah, terminal gives other position...
> With my actual project, I got error message; wanted string, got
> blaablaaClass instead.
> And that error message comes from function (from other module without Fast)
> which takes two objects are parameters.
>
Yeah, terminal gives other position...
With my actual project, I got error message; wanted string, got
blaablaaClass instead.
And that error message comes from function (from other module without Fast)
which takes two objects are parameters.
So either terminal isn't showing the real stack, or ther
The IDE is strange...
If I run gbx3 from the terminal, a correct stack backtrace is shown. The
IDE seems to miss a stack frame. I don't know why...
/Emil
2012/6/17 Jussi Lahtinen
> Aaa... OK, sorry for the trouble.
> I assumed it is because of extern,
> because with my actual project I had simi
Aaa... OK, sorry for the trouble.
I assumed it is because of extern,
because with my actual project I had similar error messages from modules
where extern is used.
Also the error message is pretty misleading as it points to the line I
copy&pasted.
I will investigate more...
Jussi
On 17 June 2
That has nothing to do with extern ;)
The code you are trying to execute is "Try oo = [TestClass1]".
As that code is always illegal, the jit throws an error directly at compile
time, before running the code. That should help people debugging code
because if there is an error in an uncommon code pa
>From this (integer to integer):
Dim iResult As Integer = mTest.DoTests()
Type mismatch: wanted Standard type, got Class instead in FMain:9.
Jussi
On 17 June 2012 02:47, Emil Lenngren wrote:
> Hi. What error do you get? The "extern" part seems to work fine for me...
>
> /Emil
>
> 2012/6/16
Hi. What error do you get? The "extern" part seems to work fine for me...
/Emil
2012/6/16 Jussi Lahtinen
> See attachment.
> If you comment out "Fast" from mTest, all works perfectly.
>
> Gambas 3 rev 4837 @ Xubuntu 12.04 64bit
>
> Jussi
>
>
> ---
See attachment.
If you comment out "Fast" from mTest, all works perfectly.
Gambas 3 rev 4837 @ Xubuntu 12.04 64bit
Jussi
GambasTester-0.9.28.tar.gz
Description: GNU Zip compressed data
--
Live Security Virtual Conferenc
36 matches
Mail list logo
