httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@worldgate.com>
Subject Re: APACHE_RELEASE no longer always increasing?
Date Sun, 07 Jun 1998 01:09:48 GMT
On Sat, 6 Jun 1998, Jim Jagielski wrote:

> This is how it's defined in httpd.h:
> 
> 
> /* Numeric release version identifier: major minor bugfix betaseq
>  * Always increases along the same track as the source branch.
>  */
> #define APACHE_RELEASE 1030101
> 
> Note that it says 'betaseq' which to my adled mind means "beta
> sequence". It does not say something like "release sequence"
> or "beta number if beta but if not beta then 1plus whatever the
> last beta was."

Note that it says it _ALWAYS_ increases.  Obviously Roy intended for
betaseq to go up one for the release, not to 00.

Even his commit message says:

------
Introduced APACHE_RELEASE numeric symbol, since branching a release
makes a date meaningless for feature comparisons.  The format is

     jnnffbb    (j = major number, nn = minor, ff = bugfix, bb = beta)

Note that 1.2b11 would have been 1020011 and 1.2.0 would be 1020012,
since the beta sequence numbers are actually X.X.0 releases.
------


> If the original discussion was different, then so what? It's kind of
> foolish to expect people to read the code and then spend time pouring
> over the archives to see if the comments in the code are correct.

Because there is zero point to HAVING APACHE_RELEASE at all if it isn't
monotonic!

It is understandable that it wasn't set correctly for 1.3 due to poor
comments in the code and the complete lack of discussion by Roy when he
introduced it on what the point is or how it works.

> 
> Now it says that it always increases along the same track as the
> source branch, but 1030000 (which is 1.3.0) IS greater than
> 1020700 (which is 1.2.7)... logically, any code that uses this
> should (1) instead use MMN if really that bothered and (2) should
> be smart enough to "ignore" the last 2 digits since they signify
> the beta sequence. The only "bits" worth notice are the first 5

The whole point for this is that you CAN'T use the MMN properly when
branches are involved.  That is the whole reason why Roy decided to add
this.

Please tell me how I can use cpp structures to do something based on the
first 5 digits.  Right now, APACHE_RELEASE is completely useless for the
purpose it was intended for, and I can't think of much else it is useful
for.

If it isn't going to be monotonic, it should be removed.


Mime
View raw message