Frederick Bruckman

Show:

What's happening?

  • Comment: Generic fcntl locking, for Solaris 10 and others

    Users of broken platforms can still use liblockfile, with its horrible sleep()'s, (or else don't keep the cache on NFS). I think there's no point in re-writing liblockfile. There's no fundamental reason why it shouldn't work. Maybe the bug is even fixed in Linux 2.6, and they just haven't updated the man pages yet?.

    2008-03-04 14:47:26 UTC in nget

  • Comment: Generic fcntl locking, for Solaris 10 and others

    That's the reason. Unless you create a ".lock" file for every newsgroup in the overview, you'd still be vulnerable to a race condition, when the group's cache is first created, or even first re-created after a flush. That's admittedly improbable in practice -- I simply didn't care for the idea of leaving the ".lock" files about for aesthetic reasons. You could skimp on lock files, but not...

    2008-03-03 16:07:07 UTC in nget

  • Comment: Patch to compile with gcc 4.2.2

    File Added: nget-gcc-4.2.2-warnings.diff.

    2007-12-05 05:45:55 UTC in nget

  • Patch to compile with gcc 4.2.2

    Here's a patch to compile with gcc 4.2.2. Also, in a separate file, a little "const" sprinkling elminates a great number of warnings.

    2007-12-05 05:43:57 UTC in nget

  • Patch to permit "strict aliasing"

    Here's a patch to avoid the "strict aliasing" warning from gcc in "rdoff.c". I simply use an assignment for a cast, but the "restrict" qualifier -- if I have this right -- permits "c" to be an alias for "data", so it should take no storage and incur no run-time overhead.

    2007-11-18 15:40:17 UTC in The Netwide Assembler

  • Comment: Without-zlib option breaks the enable-checksum=method.

    Logged In: YES user_id=803104 Does "usegz=0" in $NGETRC work for you? In general, users aren't expected to disable system libraries unless they're breaking the build, and In My Humble Opinion, it's pointless to try to make "configure" output useful and informative at every turn. What would you say to a user who doesn't have or ?.

    2006-11-13 17:12:25 UTC in nget

  • Generic fcntl locking, for Solaris 10 and others

    Here's a careful implementation of generic fcntl locking that should work on just about any system where the underlying file-system supports POSIX fcntl locks. It goes the "whole nine yards", protecting the locks by open()-ing a flag file using O_EXCL semantics. This may seem excessive at first glance, but my testing and comparison with nget's flock() locking exposed some bugs in the...

    2006-11-13 16:35:33 UTC in nget

  • A couple of patches for Solaris 10

    -- Add support for SVR4 curses. (Without this, generic_clr_bol() deletes the previous line in "screen" with default settings.) Reorganize the test, so that it checks, in turn, for "netbsd_curses", "new_curses", "old_curses". -- Allow for "pid_t" being "long" on Solaris. This merely fixes a warning on SunOS-5.10-i386, but it could address a real problem on architectures where "long"...

    2006-11-13 15:17:33 UTC in nget

About Me

  • 2003-06-17 (6 years ago)
  • 803104
  • f2bruckman (My Site)
  • Frederick Bruckman

Send me a message