On 11-Dec-2010, Osamu Aoki wrote: > In short, let me say that this documentation change caused by the > original bug report adds no value and should be reverted.
I'm not sure why there is this resistance to the change, but I will
address the salient points here briefly.
> If upstream tarball provider decides to release version less tarball,
> let's match it. Anyway, that is not going to block us. We pick latest
> version which is one with version number (null string should be oldest
> one). So the use of ".*" is not bug here.
A Debian package with an empty upstream version string violates Policy
§5.6.12:
5.6.12. `Version'
-----------------
The version number of a package. The format is:
[<epoch>`:']<upstream_version>[`-'<debian_revision>]
[…] The <upstream_version> portion of the version number is
mandatory.
Packages which use ‘(.*)’ in the ‘debian/watch’ file to match the
upstream version string have a watch file that unnecessarily allows a
Policy-violating version string to be created.
The documentation should not recommend that. That's the entire scope
of this bug report.
> Seriously, if such null matching is the REAL problem, we have many
> such REGEX in uscan code itself.
Regexes that can match the empty string are not a problem when they
are appropriate. I'm not aware of any regexes active in the ‘uscan’
code that are inappropriately matching the empty string.
The problem is simply that, before this patch, the documentation
recommended regexes that could result in an invalid Debian version
string.
None of the regexes active in the code itself have that problem,
AFAIK, which is why this bug report didn't mention them and the patch
doesn't touch them.
I hope that clears up any concerns about this fix.
--
\ “That's the essence of science: Ask an impertinent question, |
`\ and you're on the way to the pertinent answer.” —Jacob |
_o__) Bronowski, _The Ascent of Man_, 1973 |
Ben Finney <[email protected]>
signature.asc
Description: Digital signature
