Re: Re: [COMMITTERS] pgsql: Map basebackup tablespaces using a tablespace_map file

Robert Haas wrote:

> Why not?  I think that if we encounter some sort of situation that we
> think should never happen, throwing an error is exactly what we
> *should* do.  Particularly when it comes to things like removing
> files, it is very dangerous for the database to proceed if the
> situation is not as expected.  We should only remove things if we are
> quite sure that removing them is the right thing to do.

Yes, agreed.

I notice that we use %m in places where I'm not sure errno is the right
thing.  Consider the ereport() at lines 10385ff, for instance.  I don't
think fgetc() nor ferror() set errno.

I became aware of this because last week I was reading some bogus
pg_dump code that reported "could not write foobar: Success" and noticed
that the macros READ_ERROR_EXIT and WRITE_ERROR_EXIT also do
strerror(errno) after doing some fread() or similar.

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services