iswpace [was: Motif FTBFS on non-linux due to YYSTYPE]

2013-05-26 Thread Paul Gevers
Hi Graham,

[Side note: maybe we can stop using the @packages.debian.org address, I
don't think anybody else is listening.]

On 25-05-13 21:08, Graham Inggs wrote:
 Thanks for uploading!

Thank you for the nice cooperation so far, it has been a pleasure.

 I will file the bugs against the reverse dependencies of lesstif2-dev in
 Debian.

Good, but I think it is also good to wait until all architectures have
build successfully.

 I will also file one in Ubuntu to remove openmotif and replace with motif.

Good. (Same holds as above. And please let me know the bug number or
subscribe me (paul-climbing))

I believe the following is a minor thing, but I wanted to let you know
anyway. I am looking at [1], which is accessible from the top of the QA
page [2] and see that we have an interesting issue. The first item is
mentioned as an error on ia64 and powerpc:

DataF.c:4581:35: warning: array subscript is above array bounds
[-Warray-bounds]

I don't really think it is a problem at this location, but looking at
the code I see the text:
/* This routine accepts an array of wchar_t's containing wchar encodings
 * of whitespace characters (and the number of array elements), comparing
 * the wide character passed to each element of the array.  If a match
 * is found, we got a white space.  This routine exists only because
 * iswspace(3c) is not yet standard.  If a system has isw* available,
 * calls to this routine should be changed to iswspace(3c) (and callers
 * should delete initialization of the array), and this routine should
 * be deleted.  Its a stop gap measure to avoid allocating an instance
 * variable for the white_space array and/or declaring a widget wide
 * global for the data and using a macro.  Its ugly, but it works and
 * in the long run will be replaced by standard functionality. */

Maybe we should think about patching motif (and sending that upstream
obviously) to use iswspace (part of libc6). What do you think? It would
mean that motif because more locale aware with respect to whitespaces,
because now it only declares  , \t and \n, while there are more
whitespaces.

Paul

[1] https://buildd.debian.org/~brlink/packages/m/motif.html
[2] http://packages.qa.debian.org/m/motif.html



signature.asc
Description: OpenPGP digital signature


Motif FTBFS on non-linux due to YYSTYPE

2013-05-25 Thread Paul Gevers
Hi Graham,

Congratulations...

Now that we have motif in Debian/main, I would like to fix the fact that
motif FTBFS on non-linux systems. That is needed if we want to replace
lesstif2 as that is now available on more Debian platforms than motif.

I think I know where it goes wrong (I have to check to be sure) but I
don't understand why the offending check is there. The problem (I think)
is in clients/uil/UilDefI.h on line 273 and was fixed for Mac OS X in
motif bug 1479:
#if defined(linux) || defined(__APPLE__)

Do you understand why the if would be there at all? I tend to just
remove it or add a defined(__GNU__), which I think will prevent the
issue. When motif has migrated to the mirrors, I will test if this is
really the issue on some porterboxes.

If we fixed the FTBFS, do you want to start filling wishlist bugs
against the reverse dependencies of lesstif2-dev? (I am working on nedit
already). Or do you want me to start them? If you do, please don't
forget to tag them with the lesstif2motif usertag.

Paul



signature.asc
Description: OpenPGP digital signature


Re: Motif FTBFS on non-linux due to YYSTYPE

2013-05-25 Thread Paul Gevers
Hmm,

Maybe I was too quick, although I don't understand why it works now, but
kfreebsd now builds with motif 2.3.4, where it didn't build with
openmotif 2.3.3.

https://buildd.debian.org/status/package.php?p=motifsuite=sid

Paul

On 25-05-13 17:47, Paul Gevers wrote:
 Hi Graham,
 
 Congratulations...
 
 Now that we have motif in Debian/main, I would like to fix the fact that
 motif FTBFS on non-linux systems. That is needed if we want to replace
 lesstif2 as that is now available on more Debian platforms than motif.
 
 I think I know where it goes wrong (I have to check to be sure) but I
 don't understand why the offending check is there. The problem (I think)
 is in clients/uil/UilDefI.h on line 273 and was fixed for Mac OS X in
 motif bug 1479:
 #if defined(linux) || defined(__APPLE__)
 
 Do you understand why the if would be there at all? I tend to just
 remove it or add a defined(__GNU__), which I think will prevent the
 issue. When motif has migrated to the mirrors, I will test if this is
 really the issue on some porterboxes.
 
 If we fixed the FTBFS, do you want to start filling wishlist bugs
 against the reverse dependencies of lesstif2-dev? (I am working on nedit
 already). Or do you want me to start them? If you do, please don't
 forget to tag them with the lesstif2motif usertag.
 
 Paul
 




signature.asc
Description: OpenPGP digital signature


Re: Motif FTBFS on non-linux due to YYSTYPE

2013-05-25 Thread Graham Inggs
Hi Paul

Thanks for uploading!

On 25 May 2013 17:47, Paul Gevers elb...@debian.org wrote:

 If we fixed the FTBFS, do you want to start filling wishlist bugs
 against the reverse dependencies of lesstif2-dev? (I am working on nedit
 already). Or do you want me to start them? If you do, please don't
 forget to tag them with the lesstif2motif usertag.


I will file the bugs against the reverse dependencies of lesstif2-dev in
Debian.
I will also file one in Ubuntu to remove openmotif and replace with motif.

Regards
Graham