Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 26 Sep 2002 13:03:13 -0400
From:      Antoine Beaupre <anarcat@anarcat.ath.cx>
To:        Alexander Langer <alex@big.endian.de>
Cc:        Alexander Langer <alex@rtp1.SlowBlink.Com>, freebsd-libh@FreeBSD.ORG
Subject:   Package versionning (was: cvs commit: libh/lib/sysinstall Feature.cc)
Message-ID:  <D808D41E-D171-11D6-949F-0050E4A0BB3F@anarcat.ath.cx>
In-Reply-To: <20020926155153.GC47655@fump.kawo2.rwth-aachen.de>

next in thread | previous in thread | raw e-mail | index | archive | help

On Thursday, September 26, 2002, at 11:51 AM, Alexander Langer wrote:

> Thus spake Antoine Beaupre (anarcat@anarcat.ath.cx):
>
> http://docs.freebsd.org/cgi/getmsg.cgi?fetch=87661+0+archive/2002/freebsd-libh/
> 20020609.freebsd-libh
>
> That's a great deal - the portversion is only used as symbolics to ppl
> can differentiate it, and we only use the ports' serial for upgrade/undo
> checks.  I'm in favor of it :)

Hmm.. What I did is the opposite: i made the Feature scheme as to allow 
people to use *only* the regular version to record package versionning, 
so that people don't have to use cryptic serial numbers or snapshot 
dates to differentiate between their package versions.

Also, if the regular version is purely advisory, people are going to be 
*really* confused when they'll see that

ls-1.1 > ls-1.2

because their serial numbers are different. The idea is to try to avoid 
these kind of situations by making the serial number an exception rather 
than the rule: "don't use it unless you have to. You *can* rely on 
version checking or snapshot date, in some cases. But only in desperate 
cases should you fall back to using serial numbers".

In other words, the serial number is the PORTEPOCH equivalent. All 
powerful hammer, don't use it too often, because it's confusing.

Therefore, if we use regular versions for comparisons, we need a proper 
way to compare them, and we therefore need a stricter format for it.

> btw, the current version consistency check doesn't work. "1.10" is
> reported as "invalid" version.

Uh... I thought I tested that pretty thoroughly...

There are regression tests in bin/regression. Are they still passed?

I'll take a look at it tonight.

A.


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-libh" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?D808D41E-D171-11D6-949F-0050E4A0BB3F>