Re: HEADS UP: utmp.h gone. All welcome utmpx.h.

2010-01-13 Thread Pieter de Goeje
On Wednesday 13 January 2010 20:42:54 Ed Schouten wrote:
 Hello everyone,

 I just made various commits to FreeBSD HEAD to remove our old user
 accounting database interface (see utmp(5)) and replace it by the POSIX
 standardized utmpx interface (see getutxent(3)). This means we just got
 rid of some annoyances that are as old as the FreeBSD project itself:

 - Hostnames were originally restricted to 16 bytes, which is way too
   short for your average hostname generated by your ISP, but also for
   IPv6 addresses, which are at most 32 + 7 = 39 characters.

This is most welcome :-)


 - No support for login sessions not related to TTYs, like ppp(8),
   ftpd(8) sessions.

 - No support for multiple login sessions on one TTY, for example
   generated by login(1).

 I was not able to give us a smooth transition from utmp towards utmpx,
 simply because our utmp implementation offered almost no utility
 functions, which means all consumers modify the database files
 themselves. This means you should probably recompile any applications
 you're interested in that uses the user accounting database. I realize
 this may be quite uncomfortable, but we can't always win.

 [ This information is mainly for port maintainers: ]

 I've noticed there is some breakage in ports, but it shouldn't be too
 serious. I've seen cases where an application includes utmp.h, even
 though it doesn't use anything provided by that header. In other cases
 they used fields like UT_NAMESIZE to derive the maximum user name length
 supported by the system, which is clearly not what this definition was
 intended for. I've incremented __FreeBSD_version to 97 to identify
 the import of utmpx. In case a certain port breaks badly, let me know
 and I'm willing to take a look at it.

 Be sure to give it a try and report any issues. Thanks!

From the commit logs I can tell it was a lot of work on something that many 
people wouldn't consider a very glamourous or spectacular piece of FreeBSD.

So let me just say: Thanks!

Cheers,

Pieter de Goeje
___
freebsd-ports@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ports
To unsubscribe, send any mail to freebsd-ports-unsubscr...@freebsd.org


Re: ports/devel/protobuf: Segmentation fault in mmap in some applications

2010-01-07 Thread Pieter de Goeje
On Thursday 07 January 2010 10:02:36 O. Hartmann wrote:
 On 01/07/10 01:41, Pieter de Goeje wrote:
  On Wednesday 06 January 2010 14:14:28 O. Hartmann wrote:
  Dear Sirs,
  We use a software package for scientific imagery processing from USGS,
  ISIS3 (http://isis.astrogeology.usgs.gov/). The most recent version is
  3.1.21 and since this version, the software intensively uses
  libprotobuf.so.
 
  While we can use ISIS 3.1.20 very well under FreeBSD 8.0/amd64, it is
  impossible to use the software with version no. 3.1.21, which seems to
  have some issues wih libprotobuf.so. Every client out of this ISIS3
  package crashes with a segmentation fault and as far as I can judge the
  situation, there is a problem with libprotobuf.so, against which all
  clients out of ISIS 3.1.21 are linked.
 
  Perhaps the ISIS package was developed using a different (older?) version
  of Google's protocol buffers. Compiling protobuf from source is quite
  easy on FreeBSD. You can find the source here:
  http://code.google.com/p/protobuf/downloads/list
  I would start by trying version 2.1.0 and 2.2.0a.
 
  I searched for help on the ISIS3-support forum and realised that some
  Apple OS X guys have had similar problems, but those threads where
  closed immediately or got relative senseless response.
 
  In our case, we compile every necessary library and prerequisite
  software package (mostly Qt4 libs) from ports. This works great with
  some tweaks for FreeBSD in make/config.freebsd (which I derived from
  some linux and/or OS X config files).
 
  Now I'm floating like a dead man i the water. Below I provide q gdb
  output of the qview-client (the same is with all other clients, like
  photrim etc. for those familiar with the software package).
 
  A backtrace ('bt' at the gdb prompt) might contain more useful
  information.
 
  Additionaly, I provide a truss-output, that stops at mmap issues.
 
  Well, if someone could provide me with some advance debugging hints I
  would appreaciate them. I'm pretty sure he problem is located within the
  libprotobuf library or the way it is treated, but this is a guess of a
  non-developer.
 
  Thanks very much in advance.
  Please reply also to this email address, since I'm not subscriber of the
  list I post to.
 
  Oliver
 
  - Pieter

 Hello Pieter,

 ISIS3 utilises the very same revision of libprotobuf as FreeBSD has in
 the ports repositorium (libprotobuf.so.4.0.0, aka protobuf-2.2.0). The
 backtrace follows, it is a little bit lengthy ...

Ok, I can reproduce this locally. The cause is incorrect compiler flags. 
Basically one must use `pkg-config --cflags protobuf` to get the correct 
CFLAGS and `pkg-config --libs protobuf` for the correct libraries.

Most likely one or both of the following were missing during the 
compilation/linking of ISIS: -D_THREAD_SAFE -pthread

Regards,

Pieter
___
freebsd-ports@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ports
To unsubscribe, send any mail to freebsd-ports-unsubscr...@freebsd.org


Re: ports/devel/protobuf: Segmentation fault in mmap in some applications

2010-01-06 Thread Pieter de Goeje
On Wednesday 06 January 2010 14:14:28 O. Hartmann wrote:
 Dear Sirs,
 We use a software package for scientific imagery processing from USGS,
 ISIS3 (http://isis.astrogeology.usgs.gov/). The most recent version is
 3.1.21 and since this version, the software intensively uses
 libprotobuf.so.

 While we can use ISIS 3.1.20 very well under FreeBSD 8.0/amd64, it is
 impossible to use the software with version no. 3.1.21, which seems to
 have some issues wih libprotobuf.so. Every client out of this ISIS3
 package crashes with a segmentation fault and as far as I can judge the
 situation, there is a problem with libprotobuf.so, against which all
 clients out of ISIS 3.1.21 are linked.

Perhaps the ISIS package was developed using a different (older?) version of 
Google's protocol buffers. Compiling protobuf from source is quite easy on 
FreeBSD. You can find the source here: 
http://code.google.com/p/protobuf/downloads/list
I would start by trying version 2.1.0 and 2.2.0a.


 I searched for help on the ISIS3-support forum and realised that some
 Apple OS X guys have had similar problems, but those threads where
 closed immediately or got relative senseless response.

 In our case, we compile every necessary library and prerequisite
 software package (mostly Qt4 libs) from ports. This works great with
 some tweaks for FreeBSD in make/config.freebsd (which I derived from
 some linux and/or OS X config files).

 Now I'm floating like a dead man i the water. Below I provide q gdb
 output of the qview-client (the same is with all other clients, like
 photrim etc. for those familiar with the software package).

A backtrace ('bt' at the gdb prompt) might contain more useful information.


 Additionaly, I provide a truss-output, that stops at mmap issues.

 Well, if someone could provide me with some advance debugging hints I
 would appreaciate them. I'm pretty sure he problem is located within the
 libprotobuf library or the way it is treated, but this is a guess of a
 non-developer.

 Thanks very much in advance.
 Please reply also to this email address, since I'm not subscriber of the
 list I post to.

 Oliver

- Pieter
___
freebsd-ports@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ports
To unsubscribe, send any mail to freebsd-ports-unsubscr...@freebsd.org