On Mon, May 12, 2014 at 12:59 PM, Derick Rethans <[email protected]> wrote:

> On Mon, 12 May 2014, Ferenc Kovacs wrote:
>
> > On Mon, May 12, 2014 at 7:30 AM, Remi Collet <[email protected]>
> wrote:
> >
> > > Le 10/05/2014 18:16, PECL Announce a écrit :
> > > > The new PECL package pthreads-2.0.7 (stable) has been released at
> > > http://pecl.php.net/.
> > >
> > > Reports as 2.0.8...
> > >
> > >
> > maybe one of the three uploads with the same version had the correct
> > version info. :P
> > jokes aside:
> > I think we have so somehow increase the visibility of the version check
> > warning, or maybe we could introduce an ignore warnings checkbox to the
> > upload form, and if that's not checked, then any warning will abort the
> > release.
>
> This really ought to be checked at packaging time though. I wonder if
> it's worth extracting the extension build stuff out of PEAR to make a
> lightweight thingy that isn't a pain to even look at...
>
> And FYI, I was mostly against this check failing if my code wasn't doing
> the very tight format spec that it would check for. If it finds a value
> that mismatches, it should abort. And I thought that my latest change
> didn't change that actually.
>
> > ps: and I think it would be nice to have a warning, when somebody
> > tries to upload a version, which was already uploaded before.
>
> Yes.
>


Yeah, packaging time checks would be the best, but given that there were no
PEAR release in the last 3 years(http://pear.php.net/package/PEAR), and I
have a couple of fixes/improvements sitting in the the current master for
more than a year now (https://github.com/pear/pear-core) I'm a bit
reluctant to try to fix this problem through pear.
I also agree that it would be better to remove our dependency from pear, as
it is less and less actively maintained.

Up until now, I haven't realized that you intented to keep the error for
the packages which do define a PHP_EXTNAME_VERSION constant, I've checked
why this doesn't work, and I've found the bug:
We check the value of $pkg_version_ok, but we initialize it with a true
value, and never set it to false, so it will always pass:
http://git.php.net/?p=web/pecl.git;a=blob;f=public_html/release-upload.php;h=b7328d2d53a6ae06b77c3ad5eb6e542ba412a92a;hb=HEAD#l93

I've just pushed a fix, but I don't have the time to properly test it, so
it would be nice if somebody could confirm
http://git.php.net/?p=web/pecl.git;a=blobdiff;f=public_html/release-upload.php;h=681f14a47a5ba39f103bbc9b826847e072255f91;hp=b7328d2d53a6ae06b77c3ad5eb6e542ba412a92a;hb=6b34f6f3c1bbf08004626edacdd9d68b173893fc;hpb=4fae797a4a96c1a999f6e3f838c289747afb6cd4

-- 
Ferenc Kovács
@Tyr43l - http://tyrael.hu

Reply via email to