pgsql-server/src/interfaces/libpq Makefile

momjian@postgresql.org (Bruce Momjian - CVS) writes:

Modified files:
src/interfaces/libpq: Makefile

Log message:
Add dirmod to libpq Makefile.

Why? libpq doesn't use rename, and I don't think it should be using
unlink (someone *please* tell me that that unlink in getaddrinfo_unix
is a brain-dead error...)

regards, tom lane

Tom Lane wrote:

momjian@postgresql.org (Bruce Momjian - CVS) writes:

Modified files:
src/interfaces/libpq: Makefile

Log message:
Add dirmod to libpq Makefile.

Why? libpq doesn't use rename, and I don't think it should be using

Also, keep in mind that it will effect other clients who do use
rename/unlink. The question is whether they assume Unix semantics on
those commands. I don't know the answer, but I assumed they did.

unlink (someone *please* tell me that that unlink in getaddrinfo_unix
is a brain-dead error...)

Not sure --- that's why I added it.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Tom Lane wrote:

Why? libpq doesn't use rename, and I don't think it should be using

Also, keep in mind that it will effect other clients who do use
rename/unlink. The question is whether they assume Unix semantics on
those commands. I don't know the answer, but I assumed they did.

I don't believe any of our standard clients care, and in any case libpq
is not the place to put it if they do.

unlink (someone *please* tell me that that unlink in getaddrinfo_unix
is a brain-dead error...)

Not sure --- that's why I added it.

Even if it's correct (which I doubt) it's inside #ifdef
HAVE_UNIX_SOCKETS and therefore irrelevant to Windows.

regards, tom lane

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Tom Lane wrote:

unlink (someone *please* tell me that that unlink in getaddrinfo_unix
is a brain-dead error...)

Not sure --- that's why I added it.

After further research I conclude it is at best unnecessary and at worst
dangerous. It is unnecessary because Lock_AF_UNIX() will alreay have
unlink'ed any pre-existing socket file in the postmaster-startup case.
It is dangerous because there is no reason that anyone would expect
getaddrinfo() to have destructive side-effects. Certainly the
discussion of the AI_PASSIVE flag in the Linux manpage for getaddrinfo
doesn't suggest any such thing.

I'm going to remove it.

regards, tom lane