VM 4.2.1Beta Window version ?

 
Hi,
Do you have any idea if a Window version for the VM 4.2.1 exist and where i can find it ?

Thank


--
Arnaud Jean Baptiste

 
On Mon, Aug 24, 2009 at 12:29 PM, jean baptiste
arnaud<[hidden email]> wrote:
> Do you have any idea if a Window version for the VM 4.2.1 exist and where i can find it ?

I think, this numbering scheme only applies to the Mac OS VM. As far
as I know, the latest VM for windows is
http://squeakvm.org/win32/release/SqueakVM-Win32-3.11.3-bin.zip

--
Damien Cassou
http://damiencassou.seasidehosting.st

"Lambdas are relegated to relative obscurity until Java makes them
popular by not having them." James Iry

 

On 24.08.2009, at 12:29, jean baptiste arnaud wrote:

> Hi,
> Do you have any idea if a Window version for the VM 4.2.1 exist and  
> where i can find it ?


The equivalent is the latest at

http://squeakvm.org/win32/

Release numbers do not mean the same between platforms. The Windows  
3.11.3 VM is pretty much equivalent the 4.2.1 Mac VM, as is the unix  
3.10-6 VM.

<nag>except for resource directory support in the security plugin</
nag> ;)

In general, use the latest released VM for your platform.

- Bert -

 
At Mon, 24 Aug 2009 12:41:33 +0200,
Bert Freudenberg wrote:
>
> <nag>except for resource directory support in the security plugin</
> nag> ;)

  Well, so I ended up volunteering.  I don't claim that attached patch
is complete (it is working for me but security patch rewrite shouldn't
be done over night), but more reviewing and testing is really
appreciated.

  Actually, the sqWin32Security.c looks somewhat strange.  The
beginning of isAccessiblePathName() looks:

------------------------
static int isAccessiblePathName(TCHAR *pathName) {
  int i;
  /* Check if the path/file name is subdirectory of the image path */
  for(i=0; i<lstrlen(untrustedUserDirectory)-1; i++)
    if(untrustedUserDirectory[i] != pathName[i]) return 0;
------------------------

but this doesn't check the length of pathName so it can do
out-of-bounds read-access for pathName.  (it returns as soon as it
sees a char that is not part of the legitimate path , but still...)

and then it reads:
-----------
 /* special check for the trusted directory */
  if(pathName[i] == 0) return 1; /* allow access to trusted directory */
-----------

The comment is wrong, I suppose.  Not sure what is the special check
here.

And then:
-----------
 /* check last character in image path (e.g., backslash) */
  if(untrustedUserDirectory[i] != pathName[i]) return 0;
-----------
 
This, along with other lines, is a way to test the two strings are
equal.  But is the comment right?  The last character is not typically
backslash in our usual settings.

  For isAccessibleFileName(), if the function only tests the path part
anyway, it can be the same implementation to isAccessiblePathName(),
right?

expandMyDocuments() is declared as:
-----------
char *expandMyDocuments(char *pathname, char *replacement, char *result)
-----------

but the last line of the function reads:
-----------
  return strlen(result);
-----------
the caller expects the length so it should be int or size_t.

Anyway, here is my patch.  The logic is largely borrowed from the Unix
version of it by Bert.  Please review it and let us know what you
think.

-- Yoshiki

 
At Tue, 25 Aug 2009 03:30:54 -0700,
Yoshiki Ohshima wrote:
>
> Anyway, here is my patch.  The logic is largely borrowed from the Unix
> version of it by Bert.  Please review it and let us know what you
> think.

  Ah, and I sent the one without the ".." detection logic.  I'll add
it (put it back) tomorrow.

-- Yoshiki

 
Yoshiki, I had the same code and was scratching my head over it. I had  
forgotten that I had originally cloned the Windows logic.
However when I was comparing to the unix code I decided to dump that  
code base and replace it with Ian's rewrite & Bert's modifications.

You might consider reviewing the unix code for isAccessiblePathName &  
etc so all VM have the same behaviour.


On 25-Aug-09, at 3:30 AM, Yoshiki Ohshima wrote:

--
=
=
=
========================================================================
John M. McIntosh <[hidden email]>   Twitter:  
squeaker68882
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
=
=
=
========================================================================

 
At Tue, 25 Aug 2009 10:09:13 -0700,
John M McIntosh wrote:
>
> Yoshiki, I had the same code and was scratching my head over it. I had  
> forgotten that I had originally cloned the Windows logic.
> However when I was comparing to the unix code I decided to dump that  
> code base and replace it with Ian's rewrite & Bert's modifications.
>
> You might consider reviewing the unix code for isAccessiblePathName &  
> etc so all VM have the same behaviour.

  Thank you, John.  Attached is a revised patch, which could be good
to get in.

  As I wrote, the logic is almost the same as the Unix version; but
there is no realpath(), and there is no strong urge to support
symlinks.  So it only tests if the remaining part contains "\.." (not
just "..", which could be a part of legitimate file name).

-- Yoshiki

 

On 25.08.2009, at 19:36, Yoshiki Ohshima wrote:


In your patch, resourceDirectory defaults to "C:\My Squeak". I think  
the image directory is a better default? That's what I did for unix  
anyway.

- Bert -

 
At Tue, 25 Aug 2009 20:16:06 +0200,
Bert Freudenberg wrote:
>
> In your patch, resourceDirectory defaults to "C:\My Squeak". I think  
> the image directory is a better default? That's what I did for unix  
> anyway.

  Ah, yes.  I'll do that.

  I modified the installer script so it generates the right value for
"ResourceDirectory" key in the .ini file.  So installaion from this
installer, it seems to be working.

-- Yoshiki

 
Hi Yoshiki -

Can you send me the full file instead of a diff? It's easy to diff in
SVN but it's not easy to apply a patch on Windows ;-)

Cheers,
   - Andreas

Yoshiki Ohshima wrote:

 
At Tue, 25 Aug 2009 11:46:11 -0700,
Andreas Raab wrote:
>
> Hi Yoshiki -
>
> Can you send me the full file instead of a diff? It's easy to diff in
> SVN but it's not easy to apply a patch on Windows ;-)

  Ah, I just copied the diff file from my VM compiling machine this
morning and now I have trouble to apply it on the Windows machine
here.  I'll make the default value change Bert suggested anyway.  But
if somebody can just apply the patch in the meantime so that people
can take a look at it before evening, that would be great.

-- Yoshiki

 

On 25.08.2009, at 23:00, Yoshiki Ohshima wrote:
Har har. Here you go.

- Bert -

 
Okay it's in. Do you guys need a new VM build for Etoys or do you have
your own VM building setup? If you need a new one, do you just want a
clone of 3.11.3 with this fix or do you want it updated to the latest
VMMaker? Since there is currently no external pressure on these issues
it's your pick.

Cheers,
   - Andreas

Yoshiki Ohshima wrote:

 
At Tue, 25 Aug 2009 21:14:46 -0700,
Andreas Raab wrote:
>
> Okay it's in. Do you guys need a new VM build for Etoys or do you have
> your own VM building setup? If you need a new one, do you just want a
> clone of 3.11.3 with this fix or do you want it updated to the latest
> VMMaker? Since there is currently no external pressure on these issues
> it's your pick.

  Ah, thank you, Andreas.  I just made the default value change like
below or attached.

  It turned out that our schedule is pushed back a bit and we have
some weeks.  Getting the other latest stuff like SerialPlugin
update and in sync with the new VM version and have it tested with
community would be good.  So, can you update it to the VMMaker and
actually roll a new VM?  Thank you!

---------------------------------
*** sqWin32Security.c Tue Aug 25 21:46:02 2009
--- sqWin32Security.c.yoshiki Tue Aug 25 21:45:59 2009
***************
*** 246,252 ****
    lstrcpy(untrustedUserDirectory, TEXT("C:\\My Squeak\\%USERNAME%"));
 
    /* establish untrusted user directory */
!   lstrcpy(resourceDirectory, TEXT("C:\\My Squeak"));
 
    /* Look up shGetFolderPathW */
    shGetFolderPath = (void*)GetProcAddress(LoadLibrary("SHFolder.dll"),
--- 246,255 ----
    lstrcpy(untrustedUserDirectory, TEXT("C:\\My Squeak\\%USERNAME%"));
 
    /* establish untrusted user directory */
!   lstrcpy(resourceDirectory, imagePath);
!   if (resourceDirectory[lstrlen(resourceDirectory)-1] == '\\') {
!     resourceDirectory[lstrlen(resourceDirectory)-1] = 0;
!   }
 
    /* Look up shGetFolderPathW */
    shGetFolderPath = (void*)GetProcAddress(LoadLibrary("SHFolder.dll"),
---------------------------------

-- Yoshiki

 
Yoshiki Ohshima wrote:
>   It turned out that our schedule is pushed back a bit and we have
> some weeks.  Getting the other latest stuff like SerialPlugin
> update and in sync with the new VM version and have it tested with
> community would be good.  So, can you update it to the VMMaker and
> actually roll a new VM?  Thank you!

Okay, 3.11.4 is in place now. I had to fix several issues in the serial
port code which didn't even compile so I'm certain nobody has tested
that code yet. If there is anyone out there who wants to try it, please
give it a shot to see if it actually works as advertised. Other than
that the VM should work just as well as 3.11.3.

Oh, and I finally remembered to add the version info to the VM.

Cheers,
   - Andreas

 

On 26.08.2009, at 07:45, Andreas Raab wrote:

Great! Thanks!

> Oh, and I finally remembered to add the version info to the VM.


You mean "Smalltalk vmVersion" does return something sensible? Can I  
see?

- Bert -

 
On Tue, Aug 25, 2009 at 10:45:33PM -0700, Andreas Raab wrote:
>
> Okay, 3.11.4 is in place now. I had to fix several issues in the serial
> port code which didn't even compile so I'm certain nobody has tested
> that code yet. If there is anyone out there who wants to try it, please
> give it a shot to see if it actually works as advertised. Other than
> that the VM should work just as well as 3.11.3.

Actually we need confirmation that the SerialPlugin works properly
on Windows *and* on unix. Specifically, we need someone to confirm
that they can communicate with some real USB serial gadget using the
named primitives on Linux, and we need someone to confirm that the
existing serial primitive support is not broken on Windows (and
hopefully that accessing 'COM1' also works).

Severin Lemaignan has previously verified that the changes (from
Mantis 7266) worked for him on Linux, but the platform sources have
changed since then so we need to recheck it.

Dave