pdftk is unstable

[helge@…]

Since gcj34 no longer compiles on ports (Tiger, Leopard), pdftk should be removed. pdftk uses obslolete I/O mechanisms (and has other problems) on all gcc4xx versions, and although it does compile, shows erratic behaviour on all platforms (hangs, coredumps).

[jmr@…]

Assigning to maintainer.

[ryandesign@…]

Because gcj34 and gcc34 are old and don't build on our supported versions of Mac OS X, I removed support for those build mechanisms in r38135. The port still builds fine with gcc41 or gcc42.

I still find myself using pdftk on occasion, and although I have been a little concerned that there has not been a release in almost two years, it still works for me, so I don't want to remove it.

I'm not aware of the "obsolete I/O mechanisms", "other problems" or "erratic behaviour" you refer to. Can you be more specific?

[helge@…]

Replying to ryandesign@…:

Because gcj34 and gcc34 are old and don't build on our supported versions of Mac OS X, I removed support for those build mechanisms in r38135. The port still builds fine with gcc41 or gcc42. I still find myself using pdftk on occasion, and although I have been a little concerned that there has not been a release in almost two years, it still works for me, so I don't want to remove it. I'm not aware of the "obsolete I/O mechanisms", "other problems" or "erratic behaviour" you refer to. Can you be more specific?

Yez, pdftk builds, and it works most of the time, but I'm seeing the same problems that I had on Fedora9: Crashes when run from the command line and hangs when run from a (CGI or similar) script.

I'm not familiar with C++ and Java, but it is my impression (from the links below) that pdftk uses unsupported and 'unrecommended' i/o- and exception-mechanisms, which may be the reason for the problems - and the fact that pdftk support has been removed from Fedora and other distros. I have noticed from debug-traces that all I/O in pdftk takes one system call per byte, extremely inefficient and possibly part of the problem.

Here are the links I collected when trying to get pdftk running on fedora9 earlier this year (no success). I moved to a PPC macmini, which has gcj3.4 and no problems, but when I upgraded to a new dual core Intel macmini, the problems were back. As a temporary solution I am running the ppc-version of pdftk on the intel machine. Sloooow, but safe.

I'm sure you will get more from these links than I did.

 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35689  http://gcc.gnu.org/ml/java/2008-03/msg00032.html

It is interesting that we're seeing the same symptoms on Fedora and Leopard, and that the behaviour is different when run from a script (no attached terminal) and from an interactive shell. It caused me to look for quota and other resource related issues, in particular after discovering the 1 byte per I/O behaviour. I found nothing, made some changes to the C++ code in order to reduce the problem, and no progress, so I ran out of time and gave up.

[blb@…]

Cc reporter of dup #16937.

[anonymous]

Milestone Port Bugs deleted

[ryandesign@…]

Does pdftk 1.43 address the concerns of this ticket?

[helge@…]

Replying to ryandesign@…:

Does pdftk 1.43 address the concerns of this ticket?

I have not tested the new version, but I have communicated with the original creator of the software, who's now back in the loop. It is thus likely that the new version will work with newer GCC/GCJs. Will report when I have had time to check it out.