syntax error 44 : ->

Matthew Hixson <hixson@poindextrose.org> writes:

Here's a problem I have with 7.1 final that I don't see with 7.1rc4. When
I do an initdb with 7.1 I get

syntax error 44 : ->

7.1rc4 and 7.1 final are practically indistinguishable. I think you
messed up and initdb'd with something else.

The symptoms look rather like a bug we fixed back in January; is it
possible you were using a beta1 or earlier copy?

regards, tom lane

Matthew Hixson writes:

Here's a problem I have with 7.1 final that I don't see with 7.1rc4. When
I do an initdb with 7.1 I get

syntax error 44 : ->

There is an unexpected character

scrolling up the screen about 52 thousand times.

...or maybe 52000 expected characters...

on line 44 of whatever the input was.

It just so happens that global.bki, one of the input files, is exactly 44
lines long. Please check that one first. Otherwise it might happen that
the file gets corrupted while initdb is mangling it, but it's hard to tell
without some more context.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

On Thu, 19 Apr 2001, Tom Lane wrote:

Matthew Hixson <hixson@poindextrose.org> writes:

Here's a problem I have with 7.1 final that I don't see with 7.1rc4. When
I do an initdb with 7.1 I get

syntax error 44 : ->

7.1rc4 and 7.1 final are practically indistinguishable. I think you
messed up and initdb'd with something else.

That's what I thought at first too. I have done both of them three times
now. Every time 7.1 fails, and 7.1rc4 succeeds. That was also part of my
confusion in the versioning. It seemed to me that 7.1rc4 was a patch
level to 7.1 because 7.1 appears to be broken.

The symptoms look rather like a bug we fixed back in January; is it
possible you were using a beta1 or earlier copy?

I used to have 6.5.3 or some other earlier version installed on this
machine, but I'm pretty sure I successfully removed it. Maybe I still
have some cruft laying around that 7.1 sees, but 7.1rc4 does not...? Any
ideas?
-M@

On Thu, 19 Apr 2001, Peter Eisentraut wrote:

It just so happens that global.bki, one of the input files, is exactly 44
lines long. Please check that one first. Otherwise it might happen that
the file gets corrupted while initdb is mangling it, but it's hard to tell
without some more context.

Aha. Apparently my global.bki in my 7.1 archive has a bunch of binary
crap at the end of the file. global.bki in 7.1rc4 is 44 lines long, but
the one from 7.1 is 168 lines long (according to wc) and contains
thousands of repetitions of ^@.
I think I got this archive from postgresql.readysetnet.org. I'll
download it again and see if that file is still that way. If so they may
have a corrupted archive.
Thanks for your help, Peter.
-M@

Matthew Hixson writes:

Aha. Apparently my global.bki in my 7.1 archive has a bunch of binary
crap at the end of the file. global.bki in 7.1rc4 is 44 lines long, but
the one from 7.1 is 168 lines long (according to wc) and contains
thousands of repetitions of ^@.
I think I got this archive from postgresql.readysetnet.org. I'll
download it again and see if that file is still that way. If so they may
have a corrupted archive.

The global.bki file is not in the archive, it is created by
src/backend/catalog/genbki.sh. The C preprocessor, sed, and awk are major
links in that chain. If any source file is corrupted it would be one of
the header files under src/include/catalog.

Btw., there are MD5 sums of the tarballs available.

--
Peter Eisentraut peter_e@gmx.net http://funkturm.homeip.net/~peter

Okay, it looks like I must have done something weird during the first
build of 7.1 that hosed the global.bki file. I just untarred it again and
it built, installed, and initdb'd just fine.
Sorry for the noise.
-M@

On Thu, 19 Apr 2001, Peter Eisentraut wrote: