split conf-qmail in to conf-qmail-bin, conf-qmail-queue, conf-qmail-man, etc.

[alanpost]

This is a technology preview. This PR is not ready to merge.

[josuah]

I see this as a fix to #2. Reading the sources again, it looks like the running part of qmail (qmail-*) are compatible with such changes:

$ grep -e '["/]queue["/]' *.[ch] | grep -v '^hier.c:'
qmail-clean.c: if (chdir("queue") == -1) _exit(111);
qmail-qread.c: if (chdir("queue") == -1) die_chdir();
qmail-queue.c: if (chdir("queue") == -1) die(62);
qmail-send.c: while (chdir("queue") == -1)
qmail-send.c: if (chdir("queue") == -1)
qmail-tcpok.c:  if (chdir("queue/lock") == -1)
qmail-tcpok.c:    strerr_die4sys(111,FATAL,"unable to chdir to ",auto_qmail,"/queue/lock: ");
qmail-tcpok.c:    strerr_die4sys(111,FATAL,"unable to write ",auto_qmail,"/queue/lock/tcpto: ");
qmail-tcpok.c:    strerr_die4sys(111,FATAL,"unable to lock ",auto_qmail,"/queue/lock/tcpto: ");
qmail-tcpok.c:    strerr_die4sys(111,FATAL,"unable to clear ",auto_qmail,"/queue/lock/tcpto: ");
qmail-tcpto.c: if (chdir("queue/lock") == -1) die_chdir();
spawn.c: if (chdir("queue/mess") == -1) _exit(111);
tcpto.c: fdlock = open_write("queue/lock/tcpto");
tcpto.c: fd = open_read("queue/lock/tcpto");

Only chdir() to the queue, or open_{read,write} in tcpto.c accesses the queue directory directly, probably omission by the author who use chdir everywhere else, and handled properly in the changes proposed.

[josuah]

I am running on this branch of notqmail for my personnal mail server since a few months and I do not have issue anywhere.

On the other hand, I have a simplistic configuration and do not explore every feature on my setup.

But this suggest that (not)qmail get back on its feet with such changes.

[josuah]

I like the idea, but this diverged too much and applying manually on top of Master would be easier.