[Debian-med-packaging] Changes to adun.app

[Andreas Tille]

Hi Yavor,

On Wed, Jan 10, 2018 at 06:00:09PM +0200, Yavor Doganov wrote:
> [ I'm not subscribed to this list so please CC me on replies.
>   I _am_ subscribed to the adun.app PTS/tracker. ]
> 
> I've made a few changes to adun.app.git recently; most of them are
> general housekeeping and reducing the compiler warnings.  I would
> appreciate if someone (Andreas?) could review them and eventually
> upload the new version.

Thanks a lot for work on this package.

> The warnings' reduction may be considered a cosmetic issue, although
> I've lost many hours with adun.app when searching for new warnings
> while build-testing GNUstep's rdepds for the last gnustep-base
> transition recently (#879738) and for the forthcoming gnustep-gui
> transition.  It would make my life easier if the warnings are reduced.
> 
> The package certainly has more bugs now than it used to and that's due
> to changes in the GNUstep core libraries.  Some of them are subtle,
> like uncaught exceptions when loading the Test script:
> 
> 2018-01-07 19:16:38.007 UL[8362:8362] Problem posting notification:
> <NSException: 0x555555b9dc60> NAME:NSRangeException REASON:Index
> 18446744073709551615 is out of range 1 (in 'objectAtIndex:')
> INFO:{Array = ({Arguments = (""); Length = "0.0"; Name = Test; Start =
> "2018-01-07 19:07:56 +0200"; ThreadId = "aneto_20aa_20028b5e_1"; });
> Count = 1; Index = 18446744073709551615; }
> 
> The Index here is completely bogus, it should be 1.  This is an
> NSTableView notification posted when the threads table (Tools ->
> Script Manager) is updated.  Most probably the exception is due to
> NS(U)Integer type transitions that have been made in gnustep-base but
> some GUI classes were not updated.  The adun.app package may still
> have types that need to be changed, although I fixed many in 0.81-7.
> I will postpone this analysis after the gnustep-gui transition because
> there might be additional issues due to the changes in 0.25.1 (which
> was never uploaded to Debian because I discovered an ABI break [1]).
> I have the updated GNUstep core packages built locally, it's just no
> fun to debug a complex GUI app in a chroot.
> 
> The patch invalid-observer fixes an Ubuntu bug [2] that is
> reproducible on Debian, although in a different way.  This is another
> issue that's obscure and should generally lead to abort due to this
> uncaught exception:
> 
> 2018-01-06 18:37:37.463 UL[24420:24420] Problem posting notification:
> <NSException: 0x555557f72d00> NAME:NSInvalidArgumentException
> REASON:NSDataMalloc(instance) does not recognize
> outlineViewSelectionDidChange: INFO:(null)
> 
> Instead, the execution proceeds and the crash occurs at
> -[NSNotificationCenter _postAndRelease] inside the NS_DURING block...
> Completely bewildering.
> 
> The patch check-return-value is almost certainly wrong (but harmless).
> I think an exception has to be raised in most cases if the functions
> fail; I'll take a closer look to the code after the transition and
> will update the patch accordingly.
> 
> My long term plans have always been to build the plugins and
> eventually package the additional ones; the package is not very useful
> without them.  However, this requires some help from a person who
> actually understands what the program does and/or is supposed to do.
> I know absolutely nothing about molecular dynamics.  I can't even
> determine if the program's behavior is correct when the molecule file
> from the Ubuntu bug is loaded and built.

I'm somehow in a similar position as you since I'm also not a user of
this program and I'm lacking the needed background in this field.
 
> It is very unfortunate that upstream appears to be inactive.

That's true and I'm afraid at some point in time we might need to remove
the package from Debian.  Actually, I'm afraid without all your help that
point would have been reached even now...

Thanks again

      Andreas.
 
> [1] https://savannah.gnu.org/bugs/index.php?52126
> [2] https://bugs.launchpad.net/ubuntu/+source/adun.app/+bug/914750
 

-- 
http://fam-tille.de