Re: [php-maint] Freeze exception: php5
On Tue, 2010-10-05 at 15:37 -0500, Raphael Geissert wrote:
> Pure debdiff is misleading:
> debdiff -w *dsc | filterdiff -x '*/tests/*' -x '*win32*' -x '*debian/*' -x
> '*TSRM/*' -x '*configure*' -x '*README*' -x '*NEWS*' -x '*pcrelib*' -x
> '*mysqlnd*' -x '*/build/*' -x '*oci8/*' -x '*libsqlite*' -x '*pear/*' -x
> '*litespeed*' -x '*milter*' -x '*nsapi*' -x '*dsp' -x '*dsw' -x '*vcproj' -x
> '*32' -x '*vc6' -x '*date/lib/*' -x '*Zend/zend_language_scanner.c' -x
> '*Zend/zend_ini_parser.output' -x '*Zend/zend_ini_parser.c' -x '*cp5022x*' -
> x '*jisx0201*' -x '*sjis*' -x '*sapi/fpm/*' | diffstat
> 277 files changed, 4307 insertions(+), 2052 deletions(-)
> sapi/fpm: new SAPI
That's a rather quick dismissal of what appears to be completely new
functionality comprising 10,000 lines of code (+ patches and packaging
files), imho. Given that the package description even says that most
people shouldn't install it, how much of an issue would not building it
for Squeeze be?
> Not to mention that there's a regressions test suite. Talking about it:
> $ filterdiff -i '*/tests/*' -x '*oci8*' diff | diffstat | tail -n1
> 415 files changed, 8815 insertions(+), 2827 deletions(-)
> (it's the same poor "changes" metric that is used to rule out new versions, so
> I think it's fair if I include it.)
It's not a case of "ruling out" changes based purely on the diffstat; it
does provide a useful yardstick for many cases, however.
> If you want ditch the work that has been done by upstream, my fellow
> maintainers, and other contributors, go ahead. Just beware that the current
> version in testing isn't ready.
Please can we try and keep the discussion constructive? Nobody wants to
"ditch" work, but we need to balance that against the risks of including
a larger update at this stage of the process.