Support list for InnoTek LIBC ()

Knut St. Osmundsen | 15 Apr 2012 22:14

gcc 3.3.5 csd 5 / kLibC 0.6.5 released

Hi,

I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to netlabs. It can be found here:

ftp://ftp.netlabs.org/pub/gcc/

Check the 0.6.5 trac milestone for a list of fixes and features:

http://svn.netlabs.org/libc/query?group=status&milestone=libc-0.6.5

This release has not been rigorously tested, I am afraid, but I hope it will be an improvement on the previous one (csd4 / 0.6.4).

Enjoy,
knut

headers

Dave Yeo | 16 Apr 2012 08:17

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

On 04/15/12 01:14 pm, Knut St. Osmundsen wrote:
> Hi,
>
> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to netlabs. It can be found here:
>
>    ftp://ftp.netlabs.org/pub/gcc/
>
> Check the 0.6.5 trac milestone for a list of fixes and features:
>
>    http://svn.netlabs.org/libc/query?group=status&milestone=libc-0.6.5
>
> This release has not been rigorously tested, I am afraid, but I hope it will be an improvement on the
previous one (csd4 / 0.6.4).

Seems much better then 0.6.4 so far. Mozilla built with 0.6.4 silently 
failed to load SSL pages, 0.6.5 loads them fine and SeaMonkey seems 
stable, even running Flash.
Thanks for the update.
Dave

headers

James Moe | 16 Apr 2012 21:53

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released


On 04/15/2012 01:14 PM, Knut St. Osmundsen wrote:
> 
> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to
> netlabs.
> 
  Is there a symbol file (libc065.sym) somewhere? There is none in the
distribution.
  The map file, libc065.map, is not suitable for mapsym.exe to
generate a symbol file. Is there another way to do that?

--

-- 
James Moe
moe dot james at sohnen-moe dot com
520.743.3936

headers

Dave Yeo | 16 Apr 2012 23:16

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

On 04/16/12 12:53 pm, James Moe wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 04/15/2012 01:14 PM, Knut St. Osmundsen wrote:
>>
>> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to
>> netlabs.
>>
>    Is there a symbol file (libc065.sym) somewhere? There is none in the
> distribution.
>    The map file, libc065.map, is not suitable for mapsym.exe to
> generate a symbol file. Is there another way to do that?
>

There's the Perl script 
http://home.earthlink.net/~steve53/os2diags/mapsymw-0.3-20101210.zip or 
IIRC Steven also has wat2map.cmd somewhere though I can't remember the URL
Dave

headers

James Moe | 17 Apr 2012 07:01

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released


On 04/16/2012 02:16 PM, Dave Yeo wrote:
>> The map file, libc065.map, is not suitable for mapsym.exe to 
>> generate a symbol file. Is there another way to do that?
> 
> There's the Perl script 
> http://home.earthlink.net/~steve53/os2diags/mapsymw-0.3-20101210.zip
> or
> 
  Thanks. That created the symbol file I was after.

--

-- 
James Moe
moe dot james at sohnen-moe dot com
520.743.3936

headers

Steven Levine | 17 Apr 2012 19:23

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

In <4F8C8C26.1000001 <at> telus.net>, on 04/16/12
   at 02:16 PM, Dave Yeo <daveryeo <at> telus.net> said:

Hi Dave,

>http://home.earthlink.net/~steve53/os2diags/mapsymw-0.3-20101210.zip or 
>IIRC Steven also has wat2map.cmd somewhere though I can't remember the
>URL

This reminds me that I neglected to upload the latest version.

 http://home.earthlink.net/~steve53/os2diags/mapsymw-0.3-20120319.zip

It's a minor update that will only affect some users of pmdf or kdb.

There are a number of versions of mapsymw.cmd floating around.  The
version at

  http://svn.netlabs.org/repos/uniaud/OCO/trunk/tools/wat2map.cmd

works OK.  If you have Perl installed, mapsym2.pl does a better job.

Steven

--

-- 
----------------------------------------------------------------------
"Steven Levine" <steve53 <at> earthlink.net>  eCS/Warp/DIY etc.
www.scoug.com www.ecomstation.com
----------------------------------------------------------------------

headers

Dave Yeo | 29 Apr 2012 05:59

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

On 04/15/12 01:14 pm, Knut St. Osmundsen wrote:
> Hi,
>
> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to netlabs. It can be found here:
>
>    ftp://ftp.netlabs.org/pub/gcc/
>
> Check the 0.6.5 trac milestone for a list of fixes and features:
>
>    http://svn.netlabs.org/libc/query?group=status&milestone=libc-0.6.5
>
> This release has not been rigorously tested, I am afraid, but I hope it will be an improvement on the
previous one (csd4 / 0.6.4).

Hi, when someone invokes Firefox built with libc065 like this,
[quote]
I have this from one of my own programs now. I invoke like this:

  argv[0] = getenv("BROWSER");
  argv[1] = url;
  argv[2] = NULL;

  spawnvp(P_NOWAIT, argv[0], argv);
  wait(NULL);
[/quote]
Firefox tries to load www. klibc .com with the klibc bracketed by 
unprintable chars. Trace shows,
01E9 0003 | Dos32ExecPgm Entry at 10:12:23.42, Return Address =
0x1D137E68 (LIBC
165 0001:00037E68)
           | Parameter 1: PCHAR = 0x04A2684C
           | Parameter 2: LONG = 0x00000028
           | Parameter 3: ULONG = 0x00000002
           | Parameter 4: PSZ = 0x023525E0
           |   "d:\apps\mozilla\firefox\firefox.exe"
           |   "kLIBC"
           |   " http://hobbes.nmsu.edu"
           | Parameter 5: PSZ = 0x04A26C0C
           |   "WP_OBJHANDLE=167126"

His program is built with libc063 with libc063.dll being the forwarder 
dll for libc065. Using the real libc063.dll works as expected.
The complainant changed to using dosstartsession as a workaround.
Dave

headers

Knut St. Osmundsen | 30 Apr 2012 15:10

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

Hi Dave.

The problem is related to kLibC using a more unixy argument passing convention when it realizes that the new
child progress is also built for kLibC. I suspect Firefox is using DosGetInfoBlocks and pid_pchcmd to get
the command line and does its own parsing. A quick mxr search turns up
http://mxr.mozilla.org/comm-central/source/mozilla/toolkit/xre/nsNativeAppSupportOS2.cpp#429
 as circumstantial evidence.

Arguments are passed via main(int argc, char **argv[, char **envp]), programs trying to be clever and get
them any other way are on their own.

For implementation reference see: 
 http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/include/klibc/startup.h
 http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/src/lib/sys/__spawnve.c#L225
 http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/src/lib/sys/__init.c#L204

Kind Regards,
 Knut.

On Apr 29, 2012, at 5:59 AM, Dave Yeo wrote:

> On 04/15/12 01:14 pm, Knut St. Osmundsen wrote:
>> Hi,
>> 
>> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to netlabs. It can be found here:
>> 
>>   ftp://ftp.netlabs.org/pub/gcc/
>> 
>> Check the 0.6.5 trac milestone for a list of fixes and features:
>> 
>>   http://svn.netlabs.org/libc/query?group=status&milestone=libc-0.6.5
>> 
>> This release has not been rigorously tested, I am afraid, but I hope it will be an improvement on the
previous one (csd4 / 0.6.4).
> 
> Hi, when someone invokes Firefox built with libc065 like this,
> [quote]
> I have this from one of my own programs now. I invoke like this:
> 
> argv[0] = getenv("BROWSER");
> argv[1] = url;
> argv[2] = NULL;
> 
> spawnvp(P_NOWAIT, argv[0], argv);
> wait(NULL);
> [/quote]
> Firefox tries to load www. klibc .com with the klibc bracketed by unprintable chars. Trace shows,
> 01E9 0003 | Dos32ExecPgm Entry at 10:12:23.42, Return Address =
> 0x1D137E68 (LIBC
> 165 0001:00037E68)
>          | Parameter 1: PCHAR = 0x04A2684C
>          | Parameter 2: LONG = 0x00000028
>          | Parameter 3: ULONG = 0x00000002
>          | Parameter 4: PSZ = 0x023525E0
>          |   "d:\apps\mozilla\firefox\firefox.exe"
>          |   "kLIBC"
>          |   " http://hobbes.nmsu.edu"
>          | Parameter 5: PSZ = 0x04A26C0C
>          |   "WP_OBJHANDLE=167126"
> 
> His program is built with libc063 with libc063.dll being the forwarder dll for libc065. Using the real
libc063.dll works as expected.
> The complainant changed to using dosstartsession as a workaround.
> Dave
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: libc-user-unsubscribe <at> netlabs.org
> For additional commands, e-mail: libc-user-help <at> netlabs.org
> 
>

headers

Dave Yeo | 5 May 2012 04:17

Re: gcc 3.3.5 csd 5 / kLibC 0.6.5 released

On 04/30/12 06:10 am, Knut St. Osmundsen wrote:
> Hi Dave.
>
> The problem is related to kLibC using a more unixy argument passing convention when it realizes that the
new child progress is also built for kLibC. I suspect Firefox is using DosGetInfoBlocks and pid_pchcmd to
get the command line and does its own parsing. A quick mxr search turns up
http://mxr.mozilla.org/comm-central/source/mozilla/toolkit/xre/nsNativeAppSupportOS2.cpp#429
 as circumstantial evidence.
>
> Arguments are passed via main(int argc, char **argv[, char **envp]), programs trying to be clever and get
them any other way are on their own.
>
> For implementation reference see:
>   http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/include/klibc/startup.h
>   http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/src/lib/sys/__spawnve.c#L225
>   http://svn.netlabs.org/libc/browser/branches/libc-0.6/src/emx/src/lib/sys/__init.c#L204
>
> Kind Regards,
>   Knut.
>

I ended up going with the simple solution suggested by Steven, use the 
os2stub.bin from 0.6.3 to change the signature.
Thanks for the explanation,
Dave

>
>
> On Apr 29, 2012, at 5:59 AM, Dave Yeo wrote:
>
>> On 04/15/12 01:14 pm, Knut St. Osmundsen wrote:
>>> Hi,
>>>
>>> I've just uploaded the gcc 3.3.5 csd 5 / kLIBC 0.6.5 release to netlabs. It can be found here:
>>>
>>>    ftp://ftp.netlabs.org/pub/gcc/
>>>
>>> Check the 0.6.5 trac milestone for a list of fixes and features:
>>>
>>>    http://svn.netlabs.org/libc/query?group=status&milestone=libc-0.6.5
>>>
>>> This release has not been rigorously tested, I am afraid, but I hope it will be an improvement on the
previous one (csd4 / 0.6.4).
>>
>> Hi, when someone invokes Firefox built with libc065 like this,
>> [quote]
>> I have this from one of my own programs now. I invoke like this:
>>
>> argv[0] = getenv("BROWSER");
>> argv[1] = url;
>> argv[2] = NULL;
>>
>> spawnvp(P_NOWAIT, argv[0], argv);
>> wait(NULL);
>> [/quote]
>> Firefox tries to load www. klibc .com with the klibc bracketed by unprintable chars. Trace shows,
>> 01E9 0003 | Dos32ExecPgm Entry at 10:12:23.42, Return Address =
>> 0x1D137E68 (LIBC
>> 165 0001:00037E68)
>>           | Parameter 1: PCHAR = 0x04A2684C
>>           | Parameter 2: LONG = 0x00000028
>>           | Parameter 3: ULONG = 0x00000002
>>           | Parameter 4: PSZ = 0x023525E0
>>           |   "d:\apps\mozilla\firefox\firefox.exe"
>>           |   "kLIBC"
>>           |   " http://hobbes.nmsu.edu"
>>           | Parameter 5: PSZ = 0x04A26C0C
>>           |   "WP_OBJHANDLE=167126"
>>
>> His program is built with libc063 with libc063.dll being the forwarder dll for libc065. Using the real
libc063.dll works as expected.
>> The complainant changed to using dosstartsession as a workaround.
>> Dave