-
When a character device is listed in a positive transcript, and it exists in the filesystem with a different st_rdev (ie major and/or minor number), and uid, gid, or mode have also changed, fsdiff will accidentally print the transcript line twice instead of once. The enclosed patch against 1.13.0 fixes the problem. It looks like it should also apply cleanly against the git HEAD.
2009-10-28 02:31:14 UTC in Radmind
-
I brought this up on radmind-devel back in May '07, but then worked around it and never did anything about it. It's less a bug than a not-well-documented design decision. The consensus at the time seemed to be that -D could be reworked to not just change the starting point for transcript paths (which is what it does now) but to also change the locations of the certs & client directory, with...
2009-01-28 21:40:43 UTC in Radmind
-
A reasonably minor porting effort. This, along with the libsnet patch I also just posted, will allow building radmind HEAD (as of 6/7/07) on HP/UX 11.11. Client and server both have gotten some moderate testing, and Solaris 10, linux, and OSX 10.4 do not appear to be affected by the changes.
Changes that had to be made were:
- no wait4(), so emulated with wait3(). Another opt.
2007-06-08 16:03:08 UTC in Radmind
-
HP/UX and other "traditional" platforms have a version of /bin/sh that doesn't understand the == test. Since the two statements using it weren't doing any pattern matching, using a single = is equivalent and more portable.
2007-06-08 15:44:59 UTC in libsnet
-
In our environment, radmind was finding zlib, but snet was not, which led to structure length mismatches and lots of no fun. It looks like $answer was being used where $withval is correct -- at least, that's what radmind's version of the same code does, and it works for me.
2007-05-30 20:49:19 UTC in libsnet
