Обсуждение: possible problem with version string in win32 port

On a lark I changed the postgresql version string from 8.0.0beta4 to
8.0.0 in configure, did a full make clean + configure + compile.  After
this trying to start the server elicits a version mismatch complaint
from pg_ctl with postmaster.exe.  I figure probably a line ending issue,
no big deal (although I have no idea why it works now.)

So, I hack exec.c to do a straight strcmp to a line-ending insensitive
version of the same (aside: this may be a good idea anyways).  Still no
luck starting the server.  It turns out pg_ctl is trying to compare

postmaster (PostgreSQL) 8.0.0\n   <-- on the pg_ctl side
with
postmaster (PostgreSQL) 8.0.\n    <-- on the postmaster side, read from
popen with -V switch.

However, postmaster -V from the command line returns the correct version
string.

I further hack my insensitive comparison to subtract one from the
expected string length, and now everything works normally.  I checked
all the buffer lengths and everything is within normal limits.  Setting
the version string back to 8.0.0 beta4 allows everything to work as
originally coded via a straight strcmp().

Merlin

"Merlin Moncure" <merlin.moncure@rcsonline.com> writes:
> ...  It turns out pg_ctl is trying to compare

> postmaster (PostgreSQL) 8.0.0\n   <-- on the pg_ctl side
> with
> postmaster (PostgreSQL) 8.0.\n    <-- on the postmaster side, read from
> popen with -V switch.

I couldn't reproduce this on Linux.  There is Windows-specific code in
pipe_read_line(), which seems like it might be the root of the problem,
but I can't see anything obviously wrong with it.

            regards, tom lane

> "Merlin Moncure" <merlin.moncure@rcsonline.com> writes:
> > ...  It turns out pg_ctl is trying to compare
>
> > postmaster (PostgreSQL) 8.0.0\n   <-- on the pg_ctl side
> > with
> > postmaster (PostgreSQL) 8.0.\n    <-- on the postmaster side, read
from
> > popen with -V switch.
>
> I couldn't reproduce this on Linux.  There is Windows-specific code in
> pipe_read_line(), which seems like it might be the root of the
problem,
> but I can't see anything obviously wrong with it.

Me neither.  Maybe I'm just blowing smoke.  I'll double and triple check
this though...if this is a real problem it obviously has to be fixed
before the release.

Merlin