Re: [HACKERS] regression tests

Terry Mackintosh <terry(at)terrym(dot)com> writes:
> When I did make all runtest for the regression tests, I guess stderr used
> to be sent to /dev/null? but now it goes to the screen with *LOTS* of
> ERROR this and ERROR that ....

Doesn't act that way for me ... all I see is one line per test with
"ok" or "failed", same as it ever was.

> but when all was done I looked in the
> output file and the only thing that failed was triggers, and a diff of
> expected/results follows:

> [postgres(at)laptop /usr/src/pgsql/src/test/regress]$ diff
> expected/triggers.out results/triggers.out
> 53d52
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted
> 56,57d54
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted
> 59d55
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted
> 62,63d57
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted
> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted
> [postgres(at)laptop /usr/src/pgsql/src/test/regress]$

> So ... is this OK?

No. Something's wrong.

As far as the error messages to screen go, maybe someone changed the
regression driver shell script in a way that doesn't work on your shell.
Can you check against the last driver script that behaved properly for
you?

Dunno about the triggers diffs, but most other people have reported
an exact match for that one.

regards, tom lane