On Nov 2, 2010, at 5:28 AM, Michael Schroeder wrote:
> On Mon, Nov 01, 2010 at 04:50:51PM -0400, Jeff Johnson wrote:
>> Consider this perfectly acceptable (if twiddle-in-version is implemented)
>> Provides: foo = ~~~
>>
>> If that Provides: is installed in an rpmdb, the existing RPM EVR comparison
>> will see explicit existing members of the {E,V,R} triple that WILL start
>> matching with constructs like
>> Requires: foo = 1
>> and (more importantly, yo added the fix in the past week)
>> Conflicts: foo = 1
>
> Do you mean this matches now, or this will match with the changed
> algorithm? I don't see how there can be a match, the provides has
> E = 0
> V = ~~~
> R = missing
> The requires has
> E = 0
> V = 1
> R = missing
>
> "1" doesn't match "~~~".
>
I have no idea what twiddle-in-version is SUPPOSED to do in RPM.
Yes, I know what it does in Debian packaging.
The details were discussed at YOUR RPM OpenSUSE RPM Summit: No one
from @rpm5.org was invited or in attendance.
I have read the available text in Pavol's blog, @rpm.org,
and in YOUR (and 3 other) RFE reports. I listened to Pavol's
talk at FOSDEM/2011: I still haven't any clue how twiddle in version
is SUPPOSED to work.
Meanwhile I have a laundry list of 10 items from YOUR ROADMAP: 6 of them are
credibly implemented @rpm5.org.
So I'm cleaning up the remaining 4 items on YOUR ROADMAP, one of which is
twiddle in version, and closing out 3 other RFE's.
>> The open question for a twiddle-in-version implementation that MUST be
>> answered is
>>
>> How should missing values be compared to "~" placeholders meaningfully?
>>
>> (aside)
>
> IMHO the ~ support should just affect rpmEVRcmp(), i.e. the function
> used to compare the E/V/R parts.
>
Sure the implementation (if any) will be in EVRcmp.
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
