portupgrade -o strangeness...
Alex Zbyslaw
xfb52 at dial.pipex.com
Tue Jun 12 10:36:55 UTC 2007
Josh Tolbert wrote:
>On Fri, Jun 08, 2007 at 12:04:38PM +0100, Alex Zbyslaw wrote:
>
>
>>Josh Tolbert wrote:
>>
>>
>>
>>>(15:38:21 <hemi at demon:~>) $ pkg_info | grep bison
>>>bison-1.75_2,1 A parser generator from FSF, (mostly) compatible with
>>>Yacc
>>>(15:38:30 <hemi at demon:~>) $ sudo portupgrade -o devel/bison2 bison
>>>(15:38:34 <hemi at demon:~>) $ sudo portupgrade -fo devel/bison2 bison
>>>---> Reinstalling 'bison-1.75_2,1' (devel/bison)
>>>
>>>
>>>
>>>
>>Have you tried "sudo portupgrade -f -o devel/bison2 bison" in case it's
>>some bug in parsing merged options? Worth a PR in any case...
>>
>>Failing that you could just pkg_delete bison and install bison2 afresh.
>>You shouldn't have to but...
>>
>>--Alex
>>
>>
>
>Hi Alex,
>
>Yes, I did exactly that. Take a look at the example above. :)
>
>
It doesn't look like what I was suggesting is the issue so it's all
moot, but the example I can see:
sudo portupgrade -fo devel/bison2 bison
is different from what I was suggesting:
sudo portupgrade -f -o devel/bison2 bison
which deliberately split -f and -o. Your original version could reasonably be expected to work, but I have seen software (even written some :-)) which does not correctly parse flags when they are combined ("-fo") especially when one of them also takes an argument. That's not what's happening here, but my suggestion was always a shot in the dark.
>Anyway, a PR has been filed and the response is, "it's a feature." I'm not
>sure how it's a feature, but it is. The example I was given looks like this:
>
>$ pkg_version -t 2.3_1 1.75_2,1
><
>
>I'm guessing it's just doing some odd string comparison instead of breaking
>the version number apart and handling it with weight on the major version
>number, etc.
>
>
I find it bizarre too, since I don't even understand *why* the version
numbers matter in that command line. You've said "upgrade using
devel/bison2" as the origin and it's upgrading using "devel/bison". I
could understand the version number bizarre-matching affecting *whether*
portupgrade chooses to upgrade (so requiring -f) but not that it fails
to honour the origin you've given.
The pkg_version comparison is surely just wrong. The version numbers
look correct to me. Interestingly, if you drop the ,1 from the second
version, the answer is correct (on 5.4 anyway).
$ pkg_version -t 2.3_1 1.75_2
>
Or add a comma to the first
$ pkg_version -t 2.3_1,1 1.75_2,1
>
which looks like a bug to me, but maybe there's something non-standard
about that version number. Blowed if I can see what; there are plenty
of examples like it in my installed packages.
There's definitely a bug in something.
Software, bah.
--Alex
PS Presumably deinstalling bison and installing bison2 worked OK as a
workaround?
More information about the freebsd-questions
mailing list