Tcl

Steve,

At 10:26 +1100 on 8/3/01, Steve Cassidy wrote:

>I'm finding my feet with porting mac extensions. After getting the XML
>project file for QuicktimeTcl from Daniel I've managed to put together a
>project for the sample extension with only a few obvious changes. However t=
o
>get the project to link I had to add Tcl8.3.shlb and Tk8.3.shlb in addition
>to the two Stib.lib libraries.

you should not need to do that, if you do, you're locking yourself 
against a minor version of Tcl and are not taking advantage of the 
stub libraries.

>The extension now links and I can load it into tcl and the commands are the=
re
>but it crashes when I try:
>
>sha1 -string foobar
>
>so obviously something is amis.=CA
>
>  - should the project include the two shlb files?

no

>  - does someone have a port of the sample extension to CW6 and Daniel's
>build?

I was planning to do that at some stage, now might be a good a time as any.

this is the scriptics sample extension from 
ftp://ftp.ajubasolutions.com//pub/tcl/examples/tea/SampleTclExtension.tar.SI=
T.hqx
you're referring to, correct?

I'll try to update it to Pro6 and post it somewhere ASAP

>
>  my goal is not so much to get the sample extension going but to be able t=
o
>port my own code so I'd like to understand the requirements of the build
>system and why this is failing.

Cheers,

Daniel
-- 
** Daniel A. Steffen          **  "And now to something completely
** Department of Mathematics  **   different"    Monty Python
** Macquarie University       ** <mailto:steffen@...>
** NSW 2109 Australia         ** <http://www.maths.mq.edu.au/~steffen/>

At 14:09 +1100 on 8/3/01, Daniel A. Steffen wrote:

>this is the scriptics sample extension from 
>ftp://ftp.ajubasolutions.com//pub/tcl/examples/tea/SampleTclExtension.tar.SIT.hqx
>you're referring to, correct?
>
>I'll try to update it to Pro6 and post it somewhere ASAP

I've had a go at this, you can get it from

http://www.maths.mq.edu.au/~steffen/tcltk/Mac_sampleextension-0.2.1.bin

Please test and let me know if this works for you.


 From "README-MacSample":

Quick notes for 0.2.1:

* I've included an xml export of the Pro6 project, Pro5 seems to be able to
   import it and not loose access paths in the process (which is what
   happens when opening the Pro6 project in Pro5 :-( )

   Note that on Pro5 you have to add
   "{Project}:::(Pro6 Build Support):Runtime Libraries:"
   to System Access Paths ABOVE
   "{Compiler}:MacOS Support"

* I've added a Merge target which produces the final ExampleA.shlb
   extension, it merges the PPC and the CFM68k shared libraries, this allows
   for a unique binary and "package ifneeded" statement on both platforms.
   As a consequence however, the unmerged ExampleAPPC.shlb and
   ExampleACFM68K.shlb files will not load without changing their pkgIndex
   resource.

* At the moment the extension uses an .exp file to specify what symbols to
   export from its shared library, the #pragma export approach never worked
   well for me (exports too may symbols, e.g. TclStub.lib symbols).

   However, with my current patches for 8.4 TOT (which I'll hopefully get
   around to backport in time for 8.3.3), the standard Tcl DLLEXPORT macros
   now work on Mac as well, so be sure to use standard Tcl EXTERN
   declarations for symbols you want exported (i.e. for extensions, this is
   in general only your _Init (and _SafeInit if you have one) routine, see
   the Examplea_Init declaration in exampleA.h), that way you'll be able to
   switch to the "#pragma export" method and remove the .exp file with
   minimal effort in the future.


Cheers,

Daniel

-- 
** Daniel A. Steffen                  **
** <mailto:das@...> **

Daniel,
  I've succeeded in building your example extension, thanks for putting it 
together.  I've played around a little with the origanisation of it just to 
put everything in a single directory to make it easier to move to new 
extensions. I moved the project files from the Tcl build directory to a 
subdirectory of sampleextension and modified the access paths to suit. 

In the end it would seem to be best if we could distribute just the xml 
project file in the source distribution for the extension; everything can be 
recreated from this can't it? 

My next step is to try to move some of my code into here! Wish me luck.

Steve

Steve,

At 10:11 +1100 on 21/3/01, Steve Cassidy wrote:

>   I've succeeded in building your example extension, thanks for putting it
>together.  I've played around a little with the origanisation of it just to
>put everything in a single directory to make it easier to move to new
>extensions. I moved the project files from the Tcl build directory to a
>subdirectory of sampleextension and modified the access paths to suit.

that's cool,  for your own extensions modifying the access paths from 
what's suggested is obviously ok.
the build directory structure as it is now was setup by jim ingham at 
some point, it's what is suggested for widely distributed extensions 
such as tclx or itcl, to make it easy for other people to use your 
projects on a different machine (i.e. access paths to tcl and tk 
include files are relative to the project, all binaries go into build 
etc...)

>In the end it would seem to be best if we could distribute just the xml
>project file in the source distribution for the extension; everything can be
>recreated from this can't it?

hopefully, although it all depends on what MW come up with in the 
next version... there have been problems before, e.g. project files 
converted properly to a new version but not the xml etc...

it doesn't cost a lot to also include the project for the 
sampleextension, so it's maybe good to leave it there just as 
insurance...

at some point in the past I proposed to move to xml files for the Tcl 
& Tk projects in the cvs server and was met with some resistance... 
for one, the xml files sure are big...

>My next step is to try to move some of my code into here! Wish me luck.

good luck indeed! hope it works out

Cheers,

Daniel
-- 
** Daniel A. Steffen                  **
** <mailto:das@...> **

I've been working on moving some of my own code to the mac with partial 
success. I seem to have most things going for the PPC target but have a few 
remaining errors on the 68K target which block the build.  

Here's my log of what I did to port the code using Daniel's sample extension 
as an example.  It finishes off with the remaining error messages I'm 
getting. Any help in diagnosing this would be appreciated.  When it works 
I'll try to turn this into a readable HOWTO document. 

Steve


Attempt 2: based on Daniel's Sample Extension

- modified sample extension by moving project files into a subfolder of
the sampelextension folder, changed access paths in _all_ projects to
match

- unpack padgraph in the tcl/tk folder

- copy macprojects folder into padgraph, rename project to padgraph.pi

- copy MW_ExampleA.pch into MW_padgraph.pch, modify this to uncomment
#include <tk.h> and add any defines needed for the build (Codewarrior
doesn't support defines in the project/command line). Modify the names
of precompile_target near the top to reflect your project name.

- copy over exampleA.r which is the resource definition file to
padgraph.r and edit to fit your project

- copy over exampleA.exp to padgraph.expl, this defines exported
symbols, edit to fit your project

- open the project file (macprojects/padgraph.pi)

- remove the entries under Docs and replace with your own doc files 
(optional?)

- remove the entries under sources and add in your own source files
(*.c), say yes to all projects [[ I get a warning that at least one
file couldn't be added here, but they look like they were??]]

- remove the entries under headers and replace with your own

- replace MW_ExampleA.pch, exampleA.r and exampleA.exp with your new
versions in "Mac Specific Files"

- add TkStub.lib and TkStubCFM68K.lib from Tcl/Tk:Build to the Tcl
Library section

- remove shlb files from Objects (is this right??)

- click build 

- remove TkStubCFM68K.lib from the PPC project by looking under
"Link Order" and similarly remove the TkStub.lib from the C68K project. I
guess I shouldn't have added them to these projects in the first place.

- click build again

- I get:

. Warning "ignored: 'exit' in MSL MWRuntimeLibCFM68K
previously defined in MSL C.CFM68K.DLL" 
. ditto for 'abort'
. a few undefined references 'floord' 'powd' 'log10d' 'fabsd' 
. Can't find system include file <Types.r> in padgraph.r


- in the project settings for the PPC target, modify:
 - the ppc target to padgraphPPC.shlb
 - the prefix file under C/C++ Language to MW_padgraphHeaderPPC
 - in the PPC PEF section, chagen the fragment name to padgraph
 
- do the same for the 68K target and for the overall target (which
merges the two libraries into a FAT one).  

- build again, same errors

- added an explicit reference to the directory containing Types.r and
SysTypes.r in the "System Paths" of both PPC and 68K projects. Gets rid
of that error.

- all errors seem to be in the 68K version, do I need to explicitly
include the 68K math library in the project (ie. more than just
MathLib?)

here's the remaining errors:
--------------------------------
Link Warning : ignored: 'exit' in MSL MWRuntimeLibCFM68K
Previously defined in MSL C.CFM68K.DLL

Link Warning : ignored: 'abort' in MSL MWRuntimeLibCFM68K
Previously defined in MSL C.CFM68K.DLL

Link Error   : padgraph.c: 'fabsd' [XPointer] referenced from
'padgraph_modify' is undefined.

Link Error   : padgraph.c: 'log10d' [XPointer] referenced from
'padgraph_nicenum' is undefined.

Link Error   : padgraph.c: 'floord' [XPointer] referenced from
'padgraph_nicenum' is undefined.

Link Error   : padgraph.c: 'powd' [XPointer] referenced from
'padgraph_nicenum' is undefined.

Link Error   : padgraph.c: 'floord' [XPointer] referenced from
'padgraph_instancecmd' is undefined.
-----------------------------------------------

At 12:01 +1100 on 23/3/01, Steve Cassidy wrote:
>- remove the entries under sources and add in your own source files
>(*.c), say yes to all projects [[ I get a warning that at least one
>file couldn't be added here, but they look like they were??]]

note that you probably don't want to add your source files to the 
"merge" target, i.e. the target w/o  PPC or C68k in the name

>- add TkStub.lib and TkStubCFM68K.lib from Tcl/Tk:Build to the Tcl
>Library section

watch out to add these only to the appropriate targets i.e. 
TkStub.lib to PPC, TkStubCFM68K.lib to C68k

>- remove shlb files from Objects (is this right??)

no, these entries are generated automatically to match the object 
files produced by the PPC and C68k targets, to change their names, 
change names in options "PPC Target" resp "68k Target"

>- remove TkStubCFM68K.lib from the PPC project by looking under
>"Link Order" and similarly remove the TkStub.lib from the C68K project. I
>guess I shouldn't have added them to these projects in the first place.

yep, see above

>. Warning "ignored: 'exit' in MSL MWRuntimeLibCFM68K
>previously defined in MSL C.CFM68K.DLL"
>. ditto for 'abort'

normal, harmless, due to bad library design on MW's part

>. a few undefined references 'floord' 'powd' 'log10d' 'fabsd'

see below

>. Can't find system include file <Types.r> in padgraph.r
>

strange, shouldn't happen if you CW is setup properly

are you sure you have:

* system access paths include {Compiler}:MacOS Support
* your MacOS Support folder setup as follows: (the .r files are in RIncludes)
MacOS Support:
	Universal:
		Interfaces:
			AIncludes:
			CIncludes:
			ComponentIncludes:
			PInterfaces:
			RIncludes:
		Libraries:
			Classic68kLibraries:
			PPCLibraries:
			StubLibraries:


>- in the project settings for the PPC target, modify:
>  - the ppc target to padgraphPPC.shlb
>  - the prefix file under C/C++ Language to MW_padgraphHeaderPPC
>  - in the PPC PEF section, chagen the fragment name to padgraph
>

yep, see above

>- do the same for the 68K target and for the overall target (which
>merges the two libraries into a FAT one).

but given that you removed the object entries in the project, it 
probably doesn't anymore... to fix this, go into the targets panel, 
click on the disclosure triangle for your merge target and click in 
the column below the chain-link icon for both of the italicized lines 
for the PPC and 68k targets

>- added an explicit reference to the directory containing Types.r and
>SysTypes.r in the "System Paths" of both PPC and 68K projects. Gets rid
>of that error.

see above, shouldn't need to do that, but it will do the trick

>- all errors seem to be in the 68K version, do I need to explicitly
>include the 68K math library in the project (ie. more than just
>MathLib?)

no, this is due to another screwup on MW's part, I've explained the 
fix in my Tcl/Tk 8.3.2p1 distro, in the file (Pro6 Build Support):CW 
Pro6 changes, I've reproduced the relevant instructions for your 
problem below...

- in
	MacOS Support:Libraries:MacOS 
68K:MathLib68K:(Projects):MathLibCFM68K.MTarg.mcp
enable
	CFM68k->ForceIndirectAccess
for both targets and rebuild



That should do it, hope everything works now

Cheers,

Daniel
-- 
** Daniel A. Steffen          **  "And now to something completely
** Department of Mathematics  **   different"    Monty Python
** Macquarie University       ** <mailto:steffen@...>
** NSW 2109 Australia         ** <http://www.maths.mq.edu.au/~steffen/>

Almost there! 

 Thanks for the pointers Daniel. I realised a few things I'd done wrong 
mostly to do with not understanding the different columns in the project 
display (ie. that the column with the red circle indicated which files were 
part of that target -- now I understand why it says n/a on some lines).  I 
had in fact removed MathLib from the 68K target.

I had padgraph.r in the overall target which was probably why it had trouble 
finding Types.r. Removing it got rid of that problem.

I followed your instructions for rebuilding the CW library but it doesn't 
seem to have worked, now I get: Invalid PEF shared library for the 68K build 
of my project. Maybe it would have worked if I hadn't changed MathLib since 
building your sampleextension worked fine first time -- now it gets the 
Invalid PEF too.  Do I need to reinstall CW or something? 

On the good side, removing the 68K library from the final target gives me a 
library which works -- loads into wish first go and draws beautiful graph 
axes, plots data, draws a cursor. So, progress has been made.

Steve

At 0:38 +1100 on 26/3/01, Steve Cassidy wrote:

>I followed your instructions for rebuilding the CW library but it doesn't
>seem to have worked, now I get: Invalid PEF shared library for the 68K build
>of my project. Maybe it would have worked if I hadn't changed MathLib since
>building your sampleextension worked fine first time -- now it gets the
>Invalid PEF too.  Do I need to reinstall CW or something?

Steve,

I think that's a different problem, what library do you get the error 
for? I suspect ThreadsLib?
I didn't realize that would come up, c.f. an earlier post:

At 23:09 +1100 8/3/01, Daniel A. Steffen wrote:
>there was an issue where Apple stub libraries that were shipped with 
>CodeWarrior Pro6 were missing CFM68k fragments, this sounds like it. 
>Is the error message referring to ThreadsLib? then that's definitely 
>the issue you're seeing.
>
>I have reported that problem to Apple and it is fixed in the latest 
>UniversalHeaders, the solution is thus to upgrade your 
>UniversalHeaders,
>you can get recent ones from e.g.
>ftp://ftp.apple.com/developer/Development_Kits/CarbonLib_1.2.5GM_SDK.img.bin
>or from http://connect.apple.com if you are an ADC member (free BTW)
>


note that in your case I wouldn't recommend replacing the full 
universal headers with the new ones, a lot of things have changed and 
you might need to fix a number of issues with CW. See the list 
archives if you want to do this, an earlier message of mine details 
the steps needed.

I would just replace the libraries that give you trouble by the newer 
ones from the CarbonSDK

Cheers,

Daniel
-- 
** Daniel A. Steffen          **  "And now to something completely
** Department of Mathematics  **   different"    Monty Python
** Macquarie University       ** <mailto:steffen@...>
** NSW 2109 Australia         ** <http://www.maths.mq.edu.au/~steffen/>

On Monday 26 March 2001 11:16, Daniel A. Steffen wrote:
> At 0:38 +1100 on 26/3/01, Steve Cassidy wrote:
> >I followed your instructions for rebuilding the CW library but it doesn't
> >seem to have worked, now I get: Invalid PEF shared library for the 68K
> > build of my project. 
>
> I think that's a different problem, what library do you get the error
> for? I suspect ThreadsLib?
> I didn't realize that would come up, c.f. an earlier post:

No, it's for MathLib, which is why I thought the problem was with my rebuild 
of the 68K libraries.  Maybe this is still related to the problem you report 
though?

A related question: how much of a requirement is the provision of 68K 
libraries and executables these days in the Mac world? Aren't the 68K 
machines going back to the mid 90s with the equivalent of 486 processors? How 
much does providing these add to the size of distributed executables?  

Steve

At 13:19 +1000 on 26/3/01, Steve Cassidy wrote:

>No, it's for MathLib, which is why I thought the problem was with my rebuild
>of the 68K libraries.  Maybe this is still related to the problem you report
>though?

in that case, I'm not sure.
what is the exact error you're seeing (activate full error messages 
in CW's error window), and when does it happen, I presume at link 
time? of what target of what project?

>A related question: how much of a requirement is the provision of 68K
>libraries and executables these days in the Mac world? Aren't the 68K
>machines going back to the mid 90s with the equivalent of 486 processors? How
>much does providing these add to the size of distributed executables?

that's a good point, if you don't need the CFM68k executables, and 
can't test them on a 68k mac (CFM68k executables don't run on PPCs 
!), there is probably no point in spending time on a 68k version. 
however, in university environments, there are still a resonable 
number of 68k macs around... (steve, if you're interested, come by my 
office, I have a 68k mac here for testing purposes)

Cheers,

Daniel
-- 
** Daniel A. Steffen          **  "And now to something completely
** Department of Mathematics  **   different"    Monty Python
** Macquarie University       ** <mailto:steffen@...>
** NSW 2109 Australia         ** <http://www.maths.mq.edu.au/~steffen/>