Hello! Jeremie Courreges-Anglas <j...@wxcvbn.org> wrote: |here's an update to the latest release. |Test reports / comments / oks welcome.
This is great, thank you very much for your interest and efforts! v14.8+ brings in over six months of fulltime development, and is the first version of the nail/Heirloom mailx/S-nail BSD Mail codebase branch that is capable message roundtripping without causing possible message splitting due to faulty From_ line detection. v14.8 also had a Coverity.com defect density of 0.0 (fwiw). And for the first time we start to reap the fruit of all that labour: e.g., it was finally and easily possible to implement and offer a *ssl-method-HOST* variable, where HOST is the address (/port) to connect to. It was really hard to get there. Hmhm, sorry for the noise. ^.^ But now this. Please note that any v14.8 including v14.8.2 has a really severe file locking bug and should not be used -- i will release v14.8.3 by the end of the week! Optionally it is possible to use the attached patch (that was also send to the list[1]). Because i really hack many changes into v14.8.3 (the name has to be "Startled Chicken" because of that bug) without any forerun -- the opposite of the rest of v14.8! -- you may prefer this more conservative route instead? [1] http://article.gmane.org/gmane.mail.s-nail.user/276 |Note that s-nail gained a few tests since its import in the ports tree. Yes, dropping NO_TEST is due. (Tests are still a real dark corner beside that, especially expect-fail cases etc. Sigh. So to say.) Btw… Since OpenBSD ships with -lreadline, how about using WANT_READLINE=yes instead of | WANT_EDITLINE=1 since it is so much more comfortable? Otherwise the builtin command line editor may be worth looking at, but maybe without tabulator expansion (WANT_TABEXPAND=no) to not use wordexp(3). Yes, it is simple since it lets the terminal deal with the line (instead of implementing a visible viewpoint like e.g. ksh(1)), but i think doing so is still more comfortable than editline(3), especially regarding history handling. Just a suggestion. Thanks again, and Ciao! from sunny Germany right before leap, --steffen
commit 7f8ebc0 Author: Steffen (Daode) Nurpmeso <sdao...@users.sf.net> Date: 2015-06-29 13:52:34 +0200 dotlock.c:_dot_dir_access(): fix return value (Martin Neitzel).. It seems that during the rewrite i changed the control flow twice, and did not carefully enough adjust the return value accordingly, effectively setting a true value to true again instead of assigning false in the other case, and another similar case. --- dotlock.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/dotlock.c b/dotlock.c index df81fd7..43a1cbc 100644 --- a/dotlock.c +++ b/dotlock.c @@ -125,11 +125,12 @@ _dot_dir_access(char const *fname) if ((rv = is_dir(path))) { for (;;) - if (!access(path, R_OK | W_OK | X_OK)) { - rv = TRU1; + if (!access(path, R_OK | W_OK | X_OK)) break; - } else if (errno != EINTR) + else if (errno != EINTR) { + rv = FAL0; break; + } } ac_free(path); @@ -328,7 +329,7 @@ dot_lock(char const *fname, int fd, size_t pollmsecs) rv = (create_exclusive(path) == 0); olderrno = errno; sigprocmask(SIG_SETMASK, &oset, NULL); - if (!rv) + if (rv) goto jleave; while (!_dot_fcntl_lock(fd, FLOCK_UNLOCK))