headers
Martin Costabel | 1 Aug 2003 23:36
Picon

Re: Bus error on Mac OSX, again (long)

Peter O'Gorman wrote:

> I'd guess that there is a static constructor somewhere in one of the 
> other objects which gets called before main() and does something odd 

You had the right idea: As it turns out, the crash does not happen in 
the main program at all, as I thought before, but inside a static 
constructor long before main() is run. The precise line in the scribus 
code is

scribus/scfonts_encoding.cpp:23:        if (tf.open(IO_ReadOnly))

It seems that Qt's QFile::open method works differently whether it is 
run in main() or from a static constructor. Why this would be so, and 
who is to blame, Qt or Darwin, is way beyond my non-existing C++ knowledge.

As it seems, it is not a bug in Scribus, unless they are doing something 
illegal that I don't see.

I have reduced the relevant scribus code to the 2 attached small files 
that show the same behavior: If LANG is set to something non null, the 
program crashes in the tf.open command and produces the same crash log 
as scribus does, and if LANG is unset or set to the empty string, it 
doesn't crash. If instead of LANG, one plays with LC_CTYPE or LC_ALL, 
one is allowed to set them to "C", but not to anything else non empty.

Also, if the file reading code is taken out of the Foo:foo method and 
put into main(), the problem disappears. No more bus error, whether LANG 
is set or not.
(Continue reading)

Permalink | Reply | 3 comments |
headers
Yarden Livnat | 2 Aug 2003 00:29
Picon
Favicon

Panther effort


Greetings,

I've switched to Panther last week and was wondering if
I can join the effort of upgrading fink to work on Panther,
i.e., work under gcc 3.3 (I assume such an effort is under way
though I'm unable to find any info about it).

I have started my own collection of packages I compiled myself
(dlcompt, png, xemacs,...) but this is a tedious  work to do alone.

Any pointers are appreciated.

	Yarden

p.s. correct me if I'm wrong  but I assume working on Fink does not
violate the NDA on Panther, or does it ?

	the above

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
Permalink | Reply | 0 comments |
headers
Peter O'Gorman | 2 Aug 2003 01:10
Favicon
Gravatar

Re: Bus error on Mac OSX, again (long)


On Saturday, August 2, 2003, at 06:36  AM, Martin Costabel wrote:

> Peter O'Gorman wrote:
>
>> I'd guess that there is a static constructor somewhere in one of the 
>> other objects which gets called before main() and does something odd
>
> You had the right idea: As it turns out, the crash does not happen in 
> the main program at all, as I thought before, but inside a static 
> constructor long before main() is run. The precise line in the scribus 
> code is
>
> scribus/scfonts_encoding.cpp:23:        if (tf.open(IO_ReadOnly))
>
> It seems that Qt's QFile::open method works differently whether it is 
> run in main() or from a static constructor. Why this would be so, and 
> who is to blame, Qt or Darwin, is way beyond my non-existing C++ 
> knowledge.

I guess dyld hasn't bound the c++ mangled symbol "open" from Qt at this 
point, so it crashes.
I don't recall if you posted scribus's link line, is the -bind_at_load 
flag passed to ld?

Peter

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
(Continue reading)

Permalink | Reply | 1 comment |
headers
Martin Costabel | 2 Aug 2003 06:37
Picon

Re: Bus error on Mac OSX, again (long)

Peter O'Gorman wrote:
> 
> I guess dyld hasn't bound the c++ mangled symbol "open" from Qt at this 
> point, so it crashes.

Hmm, it doesn't always crash (depending on LANG), and when it crashes, 
it does so after having descended rather far down into the bowels of the 
definition of QFile::open.

> I don't recall if you posted scribus's link line, is the -bind_at_load 
> flag passed to ld?

No it doesn't, but adding it doesn't change anything.

--

-- 
Martin

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
Permalink | Reply | 0 comments |
headers
Christian Schaffner | 2 Aug 2003 17:05
Picon

Review needed: svn-ssl-0.26.0-1.info

Dear Fink Developers

In my experimental/chris01/finkinfo cvs directory there is a new 
subversion info file: svn-ssl-0.26.0-1.info
This is in preparation of a full subversion server integrated into the 
new apache2 package.

What's new (in comparison to svn-client-ssl-0.26.0-2):

- Changed name to svn-ssl to reflect that it also contains svnserver
- Made it build dynamic shared libraries
- Added -dev and -shlibs splittoffs

It would be great if someone with more experiences with -shlibs 
splittofs could have a look at it if I did it right.

Thanks a lot!
Chris.

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
Permalink | Reply | 0 comments |
headers
Benjamin Reed | 2 Aug 2003 21:27
Gravatar

Re: splitoff essential package feedback?

On Mon, 28 Jul 2003, David R. Morrison wrote:

> I'd like to get some feedback on the "splitoff" versions of the essential
> packages bzip2, gettext, libiconv, and ncurses, which were introduced to
> the unstable tree around 10 days ago.
> 
> Did the update go smoothly for people?  Is anyone aware of any problems?
> (There was one user with a problem on IRC, but I was unable to reproduce
> the problem he was having, and I've heard no other problem reports...)

Working OK here so far... I admit I haven't had that much time to play 
with it though.

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
Permalink | Reply | 7 comments |
headers
Robert Millan | 3 Aug 2003 19:09
Picon

Glibc on Darwin


[ Please keep me on CC, not subscribed ]

Hi there,

I came across your projects to build Darwin-based GNU systems, and
found them quite interesting.

Just as a suggestion to help you, I'd like to point out that Glibc
is ported (or being ported) to Darwin. For details on the port status,
refer to http://sourceforge.net/projects/glibc-darwin/

As I have experience with porting Debian/GNU userland for _both_ Glibc-based
and BSD-libc-based GNU/*BSD systems, I can tell with much reliability that
using Glibc is a _huge_ advantage over other C libraries when it comes to
porting userland like GNU and Debian that traditionaly is written with GNU
libc in mind.

When porting to BSD-libc-based, the task was real porting like what the
*BSD developers do for their systems, but when porting to GNU-libc-based,
porting was more like adding a new test case in triplet checks and trivial
things like that.

Please see this article in DP to get an idea on the advantages using Glibc
brought me for the Debian GNU/FreeBSD port:

http://www.debianplanet.org/node.php?id=968

Let me know if you have any success with Glibc-based GNU/Darwin systems.
(Continue reading)

Permalink | Reply | 2 comments |
headers
Max Horn | 3 Aug 2003 19:41
Picon
Gravatar

Re: Glibc on Darwin


Am Sonntag, 03.08.03 um 19:09 Uhr schrieb Robert Millan:

>
> [ Please keep me on CC, not subscribed ]
>
> Hi there,
>
> I came across your projects to build Darwin-based GNU systems, and
> found them quite interesting.
>
> Just as a suggestion to help you, I'd like to point out that Glibc
> is ported (or being ported) to Darwin. For details on the port status,
> refer to http://sourceforge.net/projects/glibc-darwin/

Ah yes....  0% ? :-) The project was created in 2001-10-17 21:55 , 
hasn't made any file releases, nothing in their CVS, 0 project 
activity.Did *you* actually check the port status? :-9

Anyway, to cut this reply short, so far I haven't encountered a single 
case of a software which wasn' t portable to Darwin/OSX because it had 
"Glibc in mind" when being written. Maybe in a few cases, Glibc 
provided APIs which the Darwin libc (not yet) provided, but with 10.2 
such cases are rare, and usually Glibc wouldn't be able to help it 
either, because if something isn't implemented in the Kernel/Drivers, 
then libc can only do so much to help that situation.

As such, personally I see no real advantage in porting Glibc, but maybe 
others will - this is just my two cents.
(Continue reading)

Permalink | Reply | 1 comment |
headers
Robert Millan | 4 Aug 2003 05:05
Picon

Re: Glibc on Darwin


Hi!

On Sun, Aug 03, 2003 at 07:41:19PM +0200, Max Horn wrote:
> 
> >Just as a suggestion to help you, I'd like to point out that Glibc
> >is ported (or being ported) to Darwin. For details on the port status,
> >refer to http://sourceforge.net/projects/glibc-darwin/
> 
> Ah yes....  0% ? :-) The project was created in 2001-10-17 21:55 , 
> hasn't made any file releases, nothing in their CVS, 0 project 
> activity.Did *you* actually check the port status? :-9

No, but I sent a mail to the maintainer after sending this mail to the
interested lists. My guess is that given the time that has passed some
work on sysdeps should be done (if not commited), although i have no idea
on the usability of such sysdeps extensions.

I'll let you know when i recieve some response.

Note even in case no work is done on that, I still think it's less work
and effort to write the missing sysdeps extensions than porting all of
Debian userland (which is being ported already to "generic Glibc", ie
to support Glibc with any kernel without modification)

> Anyway, to cut this reply short, so far I haven't encountered a single 
> case of a software which wasn' t portable to Darwin/OSX because it had 
> "Glibc in mind" when being written.

My comment is based on my experience with porting software between
(Continue reading)

Permalink | Reply | 0 comments |
headers
Christian Schaffner | 4 Aug 2003 15:08
Picon

Review needed: svn-ssl-0.26.0-1

Dear Fink Developers

In my experimental/chris01/finkinfo cvs directory there is a new 
subversion info file: svn-ssl-0.26.0-1.info
This is in preparation of a full subversion server integrated into the 
new apache2 package.

What's new (in comparison to svn-client-ssl-0.26.0-2):

- Changed name to svn-ssl to reflect that it also contains svnserver
- Made it build dynamic shared libraries
- Added -dev and -shlibs splittoffs

It would be great if someone with more experiences with -shlibs 
splittofs could have a look at it if I did it right.

Thanks a lot!
Chris.

-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
Permalink | Reply | 0 comments |

Gmane