[VMMaker] sqMemoryAccess.h: No such file or directory

Hi,

I have try to compile the source of the VM but I get this error:

gcc -o ./release/sqNamedPrims.o -g -mpentium -mwindows -O3 -fomit-frame-pointer -funroll-loops
-fschedule-insns2 -I. -I. -I./vm -Ic:/dx7sdk/include  -DWIN32_FILE_SUPPORT -DNO_STD_FILE_SUPPORT
-DNDEBUG
 -DLSB_FIRST -DX86 -DSQUEAK_BUILTIN_PLUGIN -c ./vm/sqNamedPrims.c
gcc -o ./release/gnu-interp.o -g -mpentium -mwindows -O3 -fomit-frame-pointer -funroll-loops
-fschedule-insns2 -I. -I. -I./vm -Ic:/dx7sdk/include  -DWIN32_FILE_SUPPORT -DNO_STD_FILE_SUPPORT
-DNDEBUG -
DLSB_FIRST -DX86 -DSQUEAK_BUILTIN_PLUGIN -c gnu-interp.c
gnu-interp.c:11: sqMemoryAccess.h: No such file or directory
make: *** [gnu-interp.o] Error 1

I use the VMMaker 3.8b6 and the win32-3.7.1 source.
I look in google but nothing came out.

Any idea?

Thanks

        Math

On Wed, 13 Sep 2006 12:11:27 +0200, Mathieu wrote:

...
> I use the VMMaker 3.8b6 and the win32-3.7.1 source.

Ah, such a combination is also interesting for me, so I wait (in the same  
queue) what comes out :)

/Klaus

>
> Math
>
>

On 13-Sep-06, at 5:17 AM, Klaus D. Witzel wrote:

> On Wed, 13 Sep 2006 12:11:27 +0200, Mathieu wrote:
>
> ...
>> I use the VMMaker 3.8b6 and the win32-3.7.1 source.
>
> Ah, such a combination is also interesting for me, so I wait (in  
> the same queue) what comes out :)

Um, guys, just think about this for a moment or two. Why would you  
expect a tool - in this case VMMaker3.8xx - to work with data clearly  
tagged - in this case win32-3.7.1 - as a different level? Surely you  
wuldn't buy 'FooBar pro for OS/X' and expect to run it on Mac OS 6 ?

Take a look at http://map1.squeakfoundation.org/package/ 
2e7f103e-22a6-470d-affe-54b1d0 and
http://map1.squeakfoundation.org/package/2e7f103e-22a6-470d- 
affe-54b1d04ef34a/autoversion/12 for the 3.8b6 version. Note that the  
first one points you to relevant pages on the swiki. Note that it  
also lists a release for Squeak 3.7 (squeak 3.7.... win32-3.7.1....  
hmmmm)

Remember that VMMaker is *not* just a tool it is a bundle of most of  
the VM source code in Slang form. It is intimately bound to a  
particular level of the platform specific code kept in the svn  
repository.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
A paperless office has about as likely as a paperless bathroom.

tim Rowledge a écrit :

Yes ok, Does the source for windows are still maintain?

Hi Tim,

on Wed, 13 Sep 2006 17:48:40 +0200, you wrote:
Please (no offense intended), what is meant by tagged, how can users like  
Mathieu or me find out the relation. Tried

- http://www.google.com/search?q=squeak+vmmaker+tagged

to no avail. The 2nd URL you posted below says "hopefully final VMMaker  
relase for building VMs to suit 3.8 and 3.9 images" and the 3.7 word and  
the win32 word does not appear on that page.

I don't know about others but when I see win32 is "only" 3.7-x it is  
automatically associated with old, outdated, "we have excluded you windoze  
users from the good stuff" (no offense intended, just feedback). BTW I  
just saw the question that Mathieu posted 10 minutes ago...

> Surely you wuldn't buy 'FooBar pro for OS/X' and expect to run it on Mac  
> OS 6 ?
>
> Take a look at http://map1.squeakfoundation.org/package/ 
> 2e7f103e-22a6-470d-affe-54b1d0 and

This one didn't work, it complained "Error when resolving package id:  
Error"

Is there a rule for "use this VMMaker version with this <platform sources>  
version". For sure the SqMap version number has nothing to do with  
platform of VM, or has it? The SqMap text for 3.8b6 does not mention win32  
or so. Perhaps this SqMap text is the source of my possible confusion (not  
the reason).

> http://map1.squeakfoundation.org/package/2e7f103e-22a6-470d- 
> affe-54b1d04ef34a/autoversion/12 for the 3.8b6 version. Note that the  
> first one points you to relevant pages on the swiki. Note that it also  
> lists a release for Squeak 3.7 (squeak 3.7.... win32-3.7.1.... hmmmm)
>
> Remember that VMMaker is *not* just a tool it is a bundle of most of the  
> VM source code in Slang form. It is intimately bound to a particular  
> level of the platform specific code kept in the svn repository.

Sure. But we unenlightened make mistakes (from which we hopefully can  
learn).

Thank you for taking care of us.

/Klaus

> tim
> --
> tim Rowledge; [hidden email]; http://www.rowledge.org/tim
> A paperless office has about as likely as a paperless bathroom.
>
>
>
>

On 13-Sep-06, at 10:09 AM, Klaus D. Witzel wrote:
[snip]
>>
>> Um, guys, just think about this for a moment or two. Why would you  
>> expect a tool - in this case VMMaker3.8xx - to work with data  
>> clearly tagged - in this case win32-3.7.1 - as a different level?
>
> Please (no offense intended), what is meant by tagged, how can  
> users like Mathieu or me find out the relation. Tried
>
> - http://www.google.com/search?q=squeak+vmmaker+tagged

OK, if I wrote something you couldn't understand then obviously I  
need to take care and be more clear.

In this case 'tagged' has no technical meaning. I was hoping that a  
file or directory named 'win32-3.7.1' would seem obviously to be  
related to a Squeak 3.7 release and therefore also related to a  
VMMaker version named 'VMMaker3.7-something'.

We have no mechanism for connecting any particular version of vmmaker  
to any particular svn version number except hoping that I remember  
(or someone reminds me) to mention it in the squeakmap page. This is  
a Bad Thing and a good simple solution would be interesting to find.

>
> to no avail. The 2nd URL you posted below says "hopefully final  
> VMMaker relase for building VMs to suit 3.8 and 3.9 images" and the  
> 3.7 word and the win32 word does not appear on that page.

That would be because vmmaker doesn't include any platform specific  
code and all the platform code is kept in the svn repository at  
http://squeakvm.org/svn/squeak .

>
> I don't know about others but when I see win32 is "only" 3.7-x it  
> is automatically associated with old, outdated, "we have excluded  
> you windoze users from the good stuff" (no offense intended, just  
> feedback). BTW I just saw the question that Mathieu posted 10  
> minutes ago...

Well that's not anything that I have any control over; Andreas looks  
after the Windows platform branch. However, in general if you want to  
try using the latest stuff yo ucan simply use svn to fetch the latest  
level of the platform code. *Most* of the time it is all up to date  
but it is most definitely possible that one or other platform can be  
out of date a little. Those of us that care for the vm don't always  
have available time at the same time and co-ordinating can be difficult.

>
>> Surely you wuldn't buy 'FooBar pro for OS/X' and expect to run it  
>> on Mac OS 6 ?
>>
>> Take a look at http://map1.squeakfoundation.org/package/ 
>> 2e7f103e-22a6-470d-affe-54b1d0 and
>
> This one didn't work, it complained "Error when resolving package  
> id: Error"

Looks like a copy/paste error, sorry about that. Try
http://map1.squeakfoundation.org/package/2e7f103e-22a6-470d- 
affe-54b1d04ef34a

>
> Is there a rule for "use this VMMaker version with this <platform  
> sources> version". For sure the SqMap version number has nothing to  
> do with platform of VM, or has it? The SqMap text for 3.8b6 does  
> not mention win32 or so. Perhaps this SqMap text is the source of  
> my possible confusion (not the reason).

I try to use a queakmap release name/number that matches the Squeak  
version (ie vmmaker3.8) plus often an internal version (ie  
vmmaker3.8b6 means vmmaker for squeak 3.8 beta 6) to indicate  
progress. Ideally the svn tree wold have matching tags at all times  
but it doesn't. It's simply a matter of available time to do it all.

[snip]
>
> Sure. But we unenlightened make mistakes (from which we hopefully  
> can learn).
Of course; and when people find problems and take the trouble to  
explain them we can improve the package and doc  to help avoid them.

Right now I need to find some time to update things to reflect the  
near-final state of squeak 3.9 and the existence of squeak 3.8.1  
which simplifies a couple of problems. Can't be soon though,  
unfortunately.


tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Useful Latin Phrases:- Anulos qui animum ostendunt omnes gestemus! =  
Let's all wear mood rings!

Klaus D. Witzel a écrit :
I coudl manage to build a vm for windows base on VMMaker-3.8b4 and win32-3.8b4 source so if you want
it ask me.

        Math

Hi Mathieu,

on Thu, 14 Sep 2006 01:35:24 +0200, you wrote:
...
> I coudl manage to build a vm for windows base on VMMaker-3.8b4 and  
> win32-3.8b4 source so if you want
> it ask me.

I do, I do :) Can I have it for making my own builds? Or is it just  
does/dont's.

Thank you in advance.

Ehm, what is it?  :)

/Klaus

> Math
>
>
>
>

Hello Mathieu,


M> I coudl manage to build a vm for windows base on VMMaker-3.8b4
M> and win32-3.8b4 source so if you want
M> it ask me.

could you please make that available somehow (or email it directly)?

I'd like to do some speed comparisons.



Thanks,


Herbert                            mailto:[hidden email]

Klaus D. Witzel a écrit :
Yes you can but it's long and you have to patch plenty of things.

1. First step you have to get the source:
 http://squeakvm.org/svn/squeak/branches/VMM38b4/
 I get it with HTTrack...

2. Load VMMaker-3.8b4 and start to generate the source.
   You will have some bug so I load some *extention from the 3.8b6 (juste some clas side categorie)
   This beceause #ccgDeclareCForVar: is missing. When you load those methods be careful of changing:
    'sqInt' -> 'int'
    'usqInt' -> 'unsigned'
       
   e.g.: ccgDeclareCForVar: aSymbolOrString

        ^'sqInt ', aSymbolOrString
-->
        ccgDeclareCForVar: aSymbolOrString

        ^'int', aSymbolOrString

  But this depend on your processor so if you have 64bit it may change.

3. Try to compile the source but you have to download the gnutools.
   You can get it from : http://www.squeakvm.org/win32/

4. When you compile you while have some parser error beceause some sqInt declaration are use.
   So replace them with the same method in 2.

5. You will also have a missing method.
   Add this at the begining of gnu-interp.c (where the build.bat are)

char *getImageName(void)
{
  return imageName;
}

  Normaly you will be able to compile at this state.
  BTW I will email the vm to you and Herbert König

Fill free to ask question (my english writing is not realy good plus I can have miss some stape)
        Math

>
> Thank you in advance.
>
> Ehm, what is it?  :)
>
> /Klaus
>
>>     Math
>>

Hi Herbert,

on Thu, 14 Sep 2006 09:59:23 +0200, you wrote:

Yes, me too. What are the areas for which you expect speed differences  
and/or verify that there are none.

/Klaus

>
> Thanks,
>
>
> Herbert                            mailto:[hidden email]
>

Hello Klaus,


KDW> Yes, me too. What are the areas for which you expect speed differences
KDW> and/or verify that there are none.
I plan to stick the vm under every 3.6 to 3.8 image I have and just
benchmark the application.

Three are about querying a mysql database and building up morpic
displays. They are written not to put put a load on the mysql
server.

One is 22 objects simulated with linear acceleration and deceleration
and "Elastischer Stoß" (elastic collision??).

One is dealing with many files in deep directory trees.

One is spending its time in making objects from ASCII files.

Try whatever else I find.

I will definitely try it on Gjallar which imho spends nearly the whole
time with seaside building html.

I'm expecting speed differences mainly by the fact that a different
toolchain might be used. I'm used to (and fed up by) looking at what
assembler code a C-compiler produces and have seen changes for the
good and for the worse with newer versions.

I'm doing this for curiosity and with the knowledge that a better
algorithm usually yields more than a faster VM.

Cheers

Herbert                            mailto:[hidden email]

Hi Herbert,

on Thu, 14 Sep 2006 12:15:27 +0200, you wrote:

> Hello Klaus,
>
>
> KDW> Yes, me too. What are the areas for which you expect speed  
> differences
> KDW> and/or verify that there are none.
>
> I plan to stick the vm under every 3.6 to 3.8 image I have and just
> benchmark the application.

...[snip] ah, application specific.$

> Try whatever else I find.
>
> I will definitely try it on Gjallar which imho spends nearly the whole
> time with seaside building html.

Yes, seaside and friends will definitely benefit even from the smallest  
performance gain; they build and GC their html, like other systems build  
and GC their ContextParts ;-)

> I'm expecting speed differences mainly by the fact that a different
> toolchain might be used. I'm used to (and fed up by) looking at what
> assembler code a C-compiler produces and have seen changes for the
> good and for the worse with newer versions.

But there is no other (tool chain for the win32 Squeak.exe series) except  
perhaps my cygwin experiment. Or are there alternatives?

> I'm doing this for curiosity and with the knowledge that a better
> algorithm usually yields more than a faster VM.

Sure, we should only use O(n+m) algo's ;-) In my experience faster is  
related to smaller hidden constants, and better is related to innovation.

/Klaus

> Cheers
>
> Herbert                            mailto:[hidden email]
>
>
>

Hi Mathieu,

thank you for taking the time and writing this up. I was able to used your  
instructions but there are some differences and enhancements, see below.

/Klaus

On Thu, 14 Sep 2006 10:48:35 +0200, you wrote:

No, there was no bug here. Downloaded #6665 full then made the upgrade to  
#6666 then loaded VMMaker again, perhaps this makes a difference.

>    This beceause #ccgDeclareCForVar: is missing. When you load those  
> methods be careful of changing:
>     'sqInt' -> 'int'
>     'usqInt' -> 'unsigned'

Did not happen here.

>    e.g.: ccgDeclareCForVar: aSymbolOrString
>
> ^'sqInt ', aSymbolOrString
> -->
> ccgDeclareCForVar: aSymbolOrString
>
> ^'int', aSymbolOrString

Again, did not happen here.

>   But this depend on your processor so if you have 64bit it may change.
>
> 3. Try to compile the source but you have to download the gnutools.
>    You can get it from : http://www.squeakvm.org/win32/

Yes, I always do.

An easier change can be made in Slang: just replace getImageName() by  
imageName in the two methods reported. So next time you don't have to  
manually patch a .c file.

>   Normaly you will be able to compile at this state.

It did and now it runs the .image from which it was generated 8-)

>   BTW I will email the vm to you and Herbert König
>
> Fill free to ask question (my english writing is not realy good plus I  
> can have miss some stape)
> Math

No problem with your English, as long as software development works :)

Thanks again.

I'm getting lost in the attributions of who said what here, so I'll  
just try to explain without quoting.

VMMaker38b4 was built in a 3.8-6665 image and should not - and  so  
far as I can see does not - have any 64 bit related changes. It most  
certainly includes
ccgDeclareCForVar: aSymbolOrString

        ^'int ', aSymbolOrString


Similarly the branch for the matching platform source on  
www.squeakvm.org -
http://www.squeakvm.org/cgi-bin/viewcvs.cgi/branches/VMM38b4/
appears to have no instances of 'sqInt' for example, so I feel  
reasonably confident it is clean.

The missing getImageName() routine is in the platform tree for the  
other platforms so it must simply be that Andreas hadn't had time to  
update code at the point where the SVN branch was made. Either work  
around is adequate pro tempore.

What really puzzles me is *why* you are attempting to build what  
amounts to a 12 month or more out of date VM? From the relevant  
emails I still have I can't see anything to suggest you tried the  
latest VMMaker-3.8b6 and the latest SVN code. I'd be a touch  
surprised if it really didn't work

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Programmer: One who is too lacking in people skills to be a software  
engineer.

Hi Tim,

on Fri, 15 Sep 2006 20:27:12 +0200, you wrote:

> I'm getting lost in the attributions of who said what here, so I'll just  
> try to explain without quoting.
>
> VMMaker38b4 was built in a 3.8-6665 image and should not - and  so far  
> as I can see does not - have any 64 bit related changes. It most  
> certainly includes
> ccgDeclareCForVar: aSymbolOrString
>
> ^'int ', aSymbolOrString

It definitively includes that. Perhaps Mathieu had something out of date  
which caused him the problem, dunno.

> Similarly the branch for the matching platform source on  
> www.squeakvm.org -
> http://www.squeakvm.org/cgi-bin/viewcvs.cgi/branches/VMM38b4/
> appears to have no instances of 'sqInt' for example, so I feel  
> reasonably confident it is clean.

It works here, does not complain about sqInt.

> The missing getImageName() routine is in the platform tree for the other  
> platforms so it must simply be that Andreas hadn't had time to update  
> code at the point where the SVN branch was made. Either work around is  
> adequate pro tempore.

I checked with the most recent Interpreter and it seemed plausible to  
replace getImageName() by imageName, on the Slang side, without a  
potential risk of crashing the VM.

> What really puzzles me is *why* you are attempting to build what amounts  
> to a 12 month or more out of date VM?

Until this point I just repeated what Mathieu did, but ...

> From the relevant emails I still have I can't see anything to suggest  
> you tried the latest VMMaker-3.8b6 and the latest SVN code.

... next the 3.8b6 VMMaker will be given a try :)

/Klaus

tim Rowledge a écrit :
So where are the 3.8b6 source for windows?

 From the relevant emails I still
> have I can't see anything to suggest you tried the latest VMMaker-3.8b6
> and the latest SVN code. I'd be a touch surprised if it really didn't work
>
> tim

On 15-Sep-06, at 1:11 PM, Mathieu wrote:

>
> So where are the 3.8b6 source for windows?
Use SVN. Get the latest code from the trunk.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Programming is an art form that fights back.

tim Rowledge a écrit :
I try and get

implicit declaration of function `sqNextEventPut' in ./vm/sqWin32DirectInput.c:250

... I use 3.8image, 3.8b6 VMMaker and the latest source in svn repository ...

Do you managed to build it Klaus?

        Math

Mathieu wrote:
> implicit declaration of function `sqNextEventPut' in ./vm/sqWin32DirectInput.c:250

Remove -ferror-implicit-function-declaration from CFLAGS. This flag is a
constant reminder for me to fix the missing prototype declarations in
the appropriate header files (one of these days I'll try to fix them all
for good).

Cheers,
   - Andreas