Thread: pg_regress starting postmaster

subject says it all. pg_regress starts "postmaster" (pg_regress.c, line
1515). Shouldn't this be "postgres" these days?

(Yes, I'm aware that I wrote that code ;-) But this just occurred to
me..)

//Magnus

> subject says it all. pg_regress starts "postmaster"
> (pg_regress.c, line 1515). Shouldn't this be "postgres" these days?
>
> (Yes, I'm aware that I wrote that code ;-) But this just occurred to
> me..)

Actually, a second thought given that I was just bitten by the
run-tests-as-admin-doesn't-work - should we use pg_ctl to start it?

//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
> subject says it all. pg_regress starts "postmaster" (pg_regress.c, line
> 1515). Shouldn't this be "postgres" these days?

No.  We're a very long way away from considering removing the
postmaster symlink, so it doesn't matter.
        regards, tom lane

"Magnus Hagander" <mha@sollentuna.net> writes:
> Actually, a second thought given that I was just bitten by the
> run-tests-as-admin-doesn't-work - should we use pg_ctl to start it?

No, not unless you'd like to break pg_regress's ability to kill the
postmaster --- we need the postmaster to be the direct child process.
        regards, tom lane

> > subject says it all. pg_regress starts "postmaster" (pg_regress.c,
> > line 1515). Shouldn't this be "postgres" these days?
>
> No.  We're a very long way away from considering removing the
> postmaster symlink, so it doesn't matter.

Well, per previous discussion, we're removing postmaster.exe from the
win32 installer, because it bloats the distribution wihtout any gain
(remember - windows doesn't have symlinks, so we need a complete copy of
a file that's 4Mb or so). So it would matter there.


> > Actually, a second thought given that I was just bitten by the
> > run-tests-as-admin-doesn't-work - should we use pg_ctl to start it?
> No, not unless you'd like to break pg_regress's ability to
> kill the postmaster --- we need the postmaster to be the
> direct child process.

D'oh, forgot about that. Nevermind about that part then.

//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
>> No.  We're a very long way away from considering removing the 
>> postmaster symlink, so it doesn't matter.

> Well, per previous discussion, we're removing postmaster.exe from the
> win32 installer, because it bloats the distribution wihtout any gain
> (remember - windows doesn't have symlinks, so we need a complete copy of
> a file that's 4Mb or so). So it would matter there.

Well, you could copy postgres.exe to postmaster.exe during install, so
I don't think you ever did need to bloat the distribution, only the
install footprint.  The question here is whether you're ready to break
existing custom scripts for starting the postmaster.  Maybe there are
none such in the wild on Windows, but I'd be hesitant to assume that.
        regards, tom lane

> >> No.  We're a very long way away from considering removing the
> >> postmaster symlink, so it doesn't matter.
>
> > Well, per previous discussion, we're removing
> postmaster.exe from the
> > win32 installer, because it bloats the distribution wihtout
> any gain
> > (remember - windows doesn't have symlinks, so we need a
> complete copy
> > of a file that's 4Mb or so). So it would matter there.
>
> Well, you could copy postgres.exe to postmaster.exe during
> install, so I don't think you ever did need to bloat the
> distribution, only the install footprint.

Except you're not supposed to do that, because the MSI auto-healing and
things like that won't work...

> The question here
> is whether you're ready to break existing custom scripts for
> starting the postmaster.  Maybe there are none such in the
> wild on Windows, but I'd be hesitant to assume that.

We're guessing there aren't - if there are, those are scripts calling
the SCM which in turns starts postgresql. So we're doing it now - if it
turns up in beta that people were actually using it from elsewhere,
we'll jus thave to put it back.

//Magnus

"Magnus Hagander" <mha@sollentuna.net> writes:
>> The question here 
>> is whether you're ready to break existing custom scripts for 
>> starting the postmaster.  Maybe there are none such in the 
>> wild on Windows, but I'd be hesitant to assume that.

> We're guessing there aren't - if there are, those are scripts calling
> the SCM which in turns starts postgresql. So we're doing it now - if it
> turns up in beta that people were actually using it from elsewhere,
> we'll jus thave to put it back.

OK.  Well, there's certainly no harm in making pg_regress execute
postgres instead of postmaster, so I'll change that.
        regards, tom lane

> >> The question here
> >> is whether you're ready to break existing custom scripts
> for starting
> >> the postmaster.  Maybe there are none such in the wild on Windows,
> >> but I'd be hesitant to assume that.
>
> > We're guessing there aren't - if there are, those are
> scripts calling
> > the SCM which in turns starts postgresql. So we're doing it
> now - if
> > it turns up in beta that people were actually using it from
> elsewhere,
> > we'll jus thave to put it back.
>
> OK.  Well, there's certainly no harm in making pg_regress
> execute postgres instead of postmaster, so I'll change that.

Thanks.

//Magnus