[pkg-horde] Bug#827483: Bug#827483: php-horde-mapi: fix autopkgtest errors

Nish Aravamudan nish.aravamudan at canonical.com
Wed Jun 29 17:56:33 UTC 2016


On 19.06.2016 [21:30:48 +0200], Mathieu Parent wrote:
> )Control: clone -1 -2
> Control: reassign -1 php-horde-mapi 1.0.8-2
> Control: tag -1 + patch
> Control: tag -2 + patch wontfix
> 
> 2016-06-19 9:46 GMT+02:00 Nish Aravamudan <nish.aravamudan at canonical.com>:
> > On 18.06.2016 [16:50:23 -0400], David Prévot wrote:
> [...]
> >> I disagree, and stand to what I’ve written in the last changelog entry:
> >>
> >>   Actually fixing the constructors requires to also fix all their calls,
> >>   both internally and externally. This backward-incompatible change has
> >>   been achieved in version 2 of phpseclib, packaged in Debian as
> >>   php-phpseclib to ensure co-installability. (Closes: #819420)
> 
> Okay. I agree with you David.
> 
> > Right, my original patch in this e-mail was just to quieten the
> > deprecated call from this testcase, as it's not really a failure (any
> > output on stderr is treated as a failure).
> >
> >> From http://phpseclib.sourceforge.net/:
> >>
> >>   The 2.0 branch has pretty much the exact same API as the 1.0 branch,
> >>   save for that it is namespaced, uses PHP5-style constructors (thereby
> >>   avoiding E_DEPRECATED errors) and requires the use of an autoloader.
> [...]
> > Agreed, and perhaps something like the attached (which passes
> > autopkgtests for me) can be applied to Debian's package (and we can sync
> > in Ubuntu) and massaged to be sent upstream? I apologize if this is way
> > off-base, I'm not all a PHP developer :)
> 
> Thanks Nish. You patch looks good.
> 
> Some notes:
> - In d/control, use php-phpseclib (>= 2~) as < and > operators are
> obsolete (see man:dpkg(1))
> - "use phpseclib\Math;" is probably not needed as you use fully
> qualified names after (see
> http://php.net/manual/en/language.namespaces.importing.php)
> - Can you send this upstream (using github.com/horde/horde or bugs.horde.org)?

I'm working on this latter part today.

Also, I just got a ping in a different upstream bug that maybe is
relevant here, about this idea:
http://cweiske.de/tagebuch/php4-constructors-php7.htm. Not sure why I
didn't think of that myself, but it would avoid the issue that led to
the regression in phpseclib, afaict, as the externally visible API would
stay the same.



More information about the pkg-horde-hackers mailing list