[Tcl-bugs] [ tktoolkit-Bugs-616245 ] tk main eats first arg

[<no...@so...>]

Bugs item #616245, was opened at 2002-09-29 19:55
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=616245&group_id=12997

Category: 74. Application Embedding
Group: 8.4.1
Status: Open
Resolution: None
Priority: 5
Submitted By: Jean-Claude Wippler (jcw)
Assigned to: Don Porter (dgp)
Summary: tk main eats first arg

Initial Comment:
When built as tclkit, tkMain.c cannot be made to ignore
the first argument.  The logic is quite broken, because
appinit gets called too late to prevent tclMain.c and
tkMain.c from abosrbing a first arg.  This required for
starpacks - which should *not* treat their first args
as scripts.  I currently get around this by repairing
the damage in my custom appinit hook - it's a bad hack
but it works.

Unfortunately, on Windows it still fails.  The fix is
quite easy: introduce the same logic in tkMain.c as is
already in tclMain.c.  I've attached the patch which
does this.

The above patch also removes the "-file" arg handling,
because that would cause starpacks to still drop a
leading "-file" arg - I hope this historical "feature"
can be dropped now.  It's not a showstopper - just a
timebomb for starpack use.

-jcw

----------------------------------------------------------------------

>Comment By: Vince Darley (vincentdarley)
Date: 2002-12-04 12:01

Message:
Logged In: YES 
user_id=32170

I agree with JCW.  Tcl_Main and Tk_Main should behave the
same wrt 'Tcl(Get|Set)StartupScriptFileName'.  Currently
they don't.

It seems as if perhaps the '-f(ile)' should stay for the
moment, but we can at least at some comments in the code
that it is obsolete and will be removed later.

Perhaps if we can get a new patch to do the above, it can be
applied quickly.



----------------------------------------------------------------------

Comment By: Don Porter (dgp)
Date: 2002-10-10 16:59

Message:
Logged In: YES 
user_id=80530

Thanks for the patch.  I think I see.  By storing
and retrieving the first argument with the
Tcl(Get|Set)StartupScriptFileName calls,
you give the AppInit routine a chance to
change whether the first argument is
used as a startup script file name.  I guess
it also gives you the chance to restore the
"argv" global variable as well.

It seems the patch leaves the usual usage
unchanged, and only adds configuration
power to the author of an AppInit routine
(even if only via internal calls).

The "-file" stuff needs to stick around for now,
probably until Tk 9.  There's a huge number of
existing wish scripts out there with "-f" embedded
in them.  I think tclkit can work around this by
processing argv[] before passing it to Tk_Main.

Is there a reason tclkit calls on Tk_Main rather
than Tcl_Main ?

----------------------------------------------------------------------

Comment By: Jeffrey Hobbs (hobbs)
Date: 2002-10-10 16:31

Message:
Logged In: YES 
user_id=72656

I do not believe that the use of -file, or -f as most people used 
it, is really fully historical yet.  I think that you will find many 
scripts doing:
#!wish -f ...
So I vote no on removing that piece.

----------------------------------------------------------------------

Comment By: Don Porter (dgp)
Date: 2002-09-30 01:06

Message:
Logged In: YES 
user_id=80530

Please attach a patch so I can be sure I understand
what it is you are requesting.  Maybe I just don't
get it.

I don't see how Tk_Main could ever be a show-stopper,
because it is only a convenience routine.  There's no
essential functionality that it exclusively controls access
to, AFAIK.  If it does what you want, great.  If not, write
your own main().  

----------------------------------------------------------------------

Comment By: Jean-Claude Wippler (jcw)
Date: 2002-09-29 22:26

Message:
Logged In: YES 
user_id=1983

Tclkit is defintiely a "wish-like" program.  It has runtime
files at the end, but it talks, looks, and quacks as a wish
- as you know.

A starpack is a tclkit with more runtime files, an is
required to act as an app.  This is *not* a recompile, but a
post-processing step (sdx -runtime).  Re-compiling is not an
option, starpacks can be generated cross-platform.

Could you please just accept the change - all I did was
include the same logic in tkMain.c as what was already
present in tclMain.c - surely this is acceptable for the
near-term?  The "don't eat -file arg" change is less
important, but this one really is a show-stopper.

I agree that longer term solutions need to address this in a
far more fundamental way - but it's hard to see how wish et
al could ever have done it right, given that they pre-date
the whole concept of starpacks.  With that perspective, the
hack that tclkit goes through to make starpacks work is
really no big deal, IMNSHO.

----------------------------------------------------------------------

Comment By: Don Porter (dgp)
Date: 2002-09-29 21:59

Message:
Logged In: YES 
user_id=80530

This is a sign that starpacks should not be calling Tk_Main
but should create their own "main" routine.

Tk_Main is for the sole purpose of creating wish-like
shell programs.  Treating the first arg as a script is
part of the definition of "wish-like".

This may be a fine idea for Tk 9, though.  Even better, I think,
would be to futher enhance the customize-ability of Tcl_Main
so that Tk_Main(Ex) can just disappear into a macro/wrapper
for Tcl_Main.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112997&aid=616245&group_id=12997