Re: running GTK

[c4t0]

Hi, thanks for replying, I'll respond it inline

On 2022-10-16 20:54, bill-auger wrote:
> without looking into this at all, there are a few remarks i
> could make
> 
> for any project which publishes versioned release tarballs, the
> VCS should be of no interest to any user - generally, the VCS
> code is of interest, only to those wishing to contribute to the
> code-base itself - so, when reporting bugs, one should always
> try to reproduce the bug with a versioned release
> 
> where gnu-smalltalk is an exception to that general rule, is
> because there has not been a stable release in a long time - the
> GTK parts of the stable release (3.2) are broken on many distros
> - arch for example, packages the "alpha" release 3.2.91, which
> works perfectly well - that really should have become the stable
> release a long time ago
> 

I use arch (btw xD) and the first thing that I tried is to install it
from the repos. When I found out that there is a gst-browser I tried to
run it and the output is:

$ gst-browser 
Object: PackageLoader class error: Invalid argument GTK: package not
found
SystemExceptions.PackageNotAvailable(Exception)>>signal
(ExcHandling.st:254)
SystemExceptions.PackageNotAvailable(Exception)>>signal:
(ExcHandling.st:264)
SystemExceptions.PackageNotAvailable class(SystemExceptions.NotFound
class)>>signalOn:what: (SysExcept.st:736)
SystemExceptions.PackageNotAvailable class>>signal: (PkgLoader.st:57)
optimized [] in PackageLoader class>>fileInPackages: (PkgLoader.st:1970)
[] in
Kernel.PackageDirectories(Kernel.PackageGroup)>>extractDependenciesFor:ifMissing:
(PkgLoader.st:165)
Kernel.PackageDirectories>>at:ifAbsent: (PkgLoader.st:231)
[] in
Kernel.PackageDirectories(Kernel.PackageGroup)>>extractDependenciesFor:ifMissing:
(PkgLoader.st:165)
Set(HashedCollection)>>do: (HashedColl.st:200)
Kernel.PackageDirectories(Kernel.PackageGroup)>>extractDependenciesFor:ifMissing:
(PkgLoader.st:173)
PackageLoader class>>fileInPackages: (PkgLoader.st:1969)
PackageLoader class>>fileInPackage: (PkgLoader.st:1959)
Package(Kernel.PackageInfo)>>fileIn (PkgLoader.st:488)
optimized [] in UndefinedObject>>executeStatements
(/usr/share/smalltalk/scripts/Load.st:126)
OrderedCollection>>do: (OrderColl.st:67)
[] in UndefinedObject>>executeStatements
(/usr/share/smalltalk/scripts/Load.st:124)
[] in BlockClosure>>ifCurtailed: (BlkClosure.st:290)
BlockClosure>>ensure: (BlkClosure.st:271)
BlockClosure>>ifCurtailed: (BlkClosure.st:276)
UndefinedObject>>executeStatements
(/usr/share/smalltalk/scripts/Load.st:114)

that's with

$ gst -v
GNU Smalltalk version 3.2.92-edb6bad6
Copyright 2009 Free Software Foundation, Inc.
Written by Steve Byrne (sbb@gnu.org) and Paolo Bonzini (bonzini@gnu.org)

GNU Smalltalk comes with NO WARRANTY, to the extent permitted by law.
You may redistribute copies of GNU Smalltalk under the terms of the
GNU General Public License.  For more information, see the file named
COPYING.

Using default kernel path: /usr/share/smalltalk/kernel
Using default image path: /var/lib/smalltalk

freshly installed

that's why I checkouted the source, to see what can I do.

I think that package mantainers simple do not compile with GTK

$ find /usr/share/smalltalk/ -iname '*.star' | grep GTK || echo not
found
not found

and

https://lists.gnu.org/archive/html/help-smalltalk/2010-05/msg00038.html

> https://git.savannah.gnu.org/cgit/smalltalk.git/snapshot/smalltalk-3.2.91.tar.gz
> https://git.savannah.gnu.org/cgit/smalltalk.git/snapshot/smalltalk-3.2.91.tar.gz.sig
> 
> that tarball corresponds to 'tag' 3.2.91 of the git sources - i
> suggest that both of you try that version, before continuing
> these threads - IMHO, any bugs reported today, should be based
> on the 3.2.91 release
> 

Just tried 'wget https://git....' and to compile it, but:

genpr-scan.l:68:5: error: redefinición de ‘yylineno’
   68 | int yylineno = 1;
      |     ^~~~~~~~
genpr-scan.c:333:5: nota: previous definition of ‘yylineno’ with type
‘int’
  333 | extern int yylineno;
      |     ^~~~~~~~

which is REALLY WEIRD 'cuz

$ echo $CFLAGS
-v -ffat-lto-objects -fPIC -Wl,--allow-multiple-definition -w

it should allow multiple definitions!

I need to do other stuff right now and not have the time to understand
what's going on... non deterministic compiling it's just to stressful...
I'll try again latter.

> actually, arch builds from
> commit=edb6bad6950c787f4b4c7f8ac5f7047ac1e8b984 ;
> but that is only because the tests in 3.2.91 fail on arch -
> they still refer to it as v3.2.91 
> 
> one other remark is that the build recipe given does not actually
> install the program - you should generally run `sudo make
> install` instead of `make install` - some programs can run
> "in-tree" but other are not designed to

yes, I know, but the error is the same either way, and I didn't want to
uninstall package repository's gst and run 'sudo make intall' every time
I touch something. How does the developers compile and run their code
when they are developing the software?

Chau!