svn commit: r370411 - head/sysutils/e2fsprogs

Matthias Andree mandree at FreeBSD.org
Wed Oct 8 06:47:47 UTC 2014


Am 08.10.2014 um 03:02 schrieb Bryan Drewery:
> On 10/7/2014 7:36 PM, Bryan Drewery wrote:
>> On 10/7/2014 6:43 PM, Matthias Andree wrote:
>>> Author: mandree
>>> Date: Tue Oct  7 23:43:46 2014
>>> New Revision: 370411
>>> URL: https://svnweb.freebsd.org/changeset/ports/370411
>>> QAT: https://qat.redports.org/buildarchive/r370411/
>>>
>>> Log:
>>>   Mark broken everywhere, upstream bug (already reported).
>>>   To avoid PORTEPOCH/rollback-and-forth-again, let's give upstream a few
>>>   days for a fix before rolling back to 1.42.10.
>>
>> This is horrible. We're providing no packages for any branch now to
>> avoid bumping a metadata counter?
>>
>>
> 
> Unfortunately poudriere does not care that this is BROKEN. It reads
> PKGVERSION, sees it is newer, deletes the old package, then finally
> finds it should IGNORE it.
> 
> A bug in poudriere sure.

Not so fast.  This is not something that poudriere could "decide" on its
own.  Actually, there is a matter of context and perhaps policy.

Consider: is the port broken but the old version is fine and the solver
of pkg finds it sufficient to go with all other packages, then leaving
the old package in place would work.
  If, however, the old version of some port were to be vulnerable, and
the current version were known to not fix the vuln', then poudriere
would have done the right thing by removing the BROKEN/FORBIDDEN/...
package.



More information about the svn-ports-all mailing list