Hello Greg,
>> Peter posted an updated version of Fabiens patch about a month ago (which at
>> this point no longer applies)
Attached a v15 which is a rebase, after some minor changes in the source
and some new test cases added (good!).
>> fixing a few issues, but also point at old review comments still
>> unaddressed.
ISTM that all comments have been addressed. However, the latest patch
raises issues about work around libpq corner case behaviors which are
really just that, corner cases.
>> Since this was pushed, but had to be reverted, I assume there is a
>> desire for the feature but it seems to need more work still.
> It looks like Peter and Fabien were debating the merits of a libpq
> change and probably that won't happen this release cycle.
ISTM these are really very minor issues that could be resolved in this
cycle.
> Is there a kernel of psql functionality that can be extracted from this
> without the libpq change in this release cycle or should it wait until
> we add the functionality to libpq?
The patch can wait for the issues to be resolved one way or an other
before proceeding, *or* it can be applied, maybe with a small tweak, and
the libpq issues be fixed separately.
For a reminder, there are two actual "issues"features" or "bug" with
libpq, which are made visible by the patch, but are pre-existing:
(1) under some circumstances a infinite stream of empty results is
returned, that has to be filtered out manually.
(2) under some special circumstances some error messages may be output
twice because of when libpq decides to reset the error buffer.
(1) has been there for ever, and (2) is under investigation to possibly
improve the situation, so as to remove a hack in the code to avoid it.
The alternative which IMO would be ok is to admit that under some very
special conditions the same error message may be output twice, and if it
is resolved later on then fine.
> If it's the latter then perhaps we should move this to 16?
I'm not that pessimistic! I may be proven wrong:-)
--
Fabien.
