Re: New bugs discovered when adding vnlog support

[Tim Rice]

Hey Erik,

> I do not see the see the "5111445" commit in git.  It is not shown
> on savannah either.

Oh yeah, I did forget to push the last couple of commits. That's now done.

Whenever you want to push something up, I don't think you need to worry about 
merge conflicts. If I forgot to push some commits, then any merge conflicts are 
on me :)


> > 4a3b349 datamash: bugfix: Allow crosstab to be called by header field name.
>
> I would like to have a test with --header-in, but without
> --header-out, because IMHO that seems like a realistic use-case,
> while creating this header line that is not really a header line,
> but rather provides some kind of description for the table, looks
> a bit dubious to me.  I would not call it wrong or useless, but
> rather a bit surprising.  Adding a few more words to explain this
> "header" might be useful.

Sure thing, makes sense. See commits 31aa0c4 and 1c76dc9.


> > 6dc46b8 tests: Add the test for pcov header in/out
>
> It might make sense to add similar tests for scov, ppearson,
> and spearson, since those should be affected by the change, too.

No worries, now in commit e6e6982.


> Do you think it would make sense to enable and use (part of) the
> tracker features of savannah for those?
>
> IMHO bug reporting via the mailing list works fine, but when we
> decide that some issues need to be addressed later or a report
> turns out to comprise several issues it might be nice to keep
> that information available on savannah.
>
> It seems to me as if the "Bug Tracker" might be appropriate:
>
> https://savannah.gnu.org/userguide/?file=ch04.html
>
> I would prefer to give write access only to group members (reports
> should go to bug-datamash@gnu.org).

Well, most of my "todo" items are not bugs. They are things like feature 
requests, refactorization ideas, reminders about some code which should be more defensive 
about memory allocation, lapses in code style that should be cleaned up, what needs doing 
before the release, and so on.

Most bugs we've dealt with in the last couple of months have been shallow 
enough that we can fix them quickly, about an hour on average. Furthermore, I 
have no expectation that these things should be triaged to you or Shawn.

So, I speculate a formal bug tracker may not have too much utility. Still, if 
you think it would be useful, I won't stand in your way. Investigating the 
savannah bugtracker is now on my todo list :)


> The man page currently states:
>
>   Some operations require field numbers (groupby, crosstab)
>   while others do not (reverse,check,transpose).
>
> With current git HEAD, both groupby and crosstab allow field
> names, so this sentence should be adjusted or deleted.

Good catch, that's fixed in 582676b.


> There is no test using groupby with a field name, I think we
> should add one.

What about test "ng11" near the bottom of `tests/datamash-tests.pl` ?

~ Tim