Re: [HACKERS] 6.4.1 schedule SUNOS4.1.X build

I isolated a code problem (sunos4.1.X does not have a fflush(NULL)),
Tom Lane is adding a portable workaround. Yes this is an ANSI required
but either forget about Sunos4.1.X or use more general code. fflush(NULL)
core dumps sunos4.1.X

Even with a non core dumping build I am calling on the group
memory to suggest what to try next. Sunos 4.1.X can be built and
postmaster started and a first psql started. On the most simple
request

\l

things just hang. With debug on I see a final

enter CommitTransactionCommand()

I added a few more diagnostic prints and CommitTransactionCommand
runs with a
case TBLOCK_DEFAULT
and says it is exiting...

Suggestions where to look next - if this a known problem let me
know where to look.
I can test fixes on both BSD/OS 3.X and BSD/OS 4.0 so I can't go too
far from an ANSI/POSIX solution before those builds will break.

The regression test also hang with the CommitTransactionCommand.

\q works and psql exits and \? gives the help list

but...

template1=> \df
Backend sent D message without prior T

and we are hung

Thanks

--
Stephen N. Kogge
stevek@uimage.com
http://www.uimage.com