On Wed, Mar 25 2020, Paco Esteban <p...@e1e0.net> wrote:
> On Sun, 15 Mar 2020, Marc Espie wrote:
>
>> On Sun, Mar 15, 2020 at 12:28:25PM +0100, Paco Esteban wrote:
>> > oh, that did not occur to me ... You mean that if port foo/bar gets
>> > EPOCH = 0, then all dependent ports should change it's dependencies
>> > declarations from foo/bar>=1.1 to foo/bar>=1.1v0 ??
>>
>> Yes, that's the main reason that led to introducing PKGSPEC in bsd.port.mk
>> and tweaking _print-metadata, so we didn't hav to change every
>> dependent port.
>>
>> You still need to bump REVISION in dependent port when you change PKGSPEC
>> though.
>
> Thanks Marc, I tried to include that too.
> All patches updated and attached again.
>
> Cheers,
Comments inline,
> Index: bsd.port.mk.5
> ===================================================================
> RCS file: /home/cvs/src/share/man/man5/bsd.port.mk.5,v
> retrieving revision 1.528
> diff -u -p -r1.528 bsd.port.mk.5
> --- bsd.port.mk.5 24 Mar 2020 14:59:52 -0000 1.528
> +++ bsd.port.mk.5 25 Mar 2020 20:07:07 -0000
> @@ -1650,6 +1650,10 @@ If set to
> will not open changed files in an editor.
> .It Ev EPOCH
> Epoch number of the current package.
> +Used when the port version is changed but the new version is not regarded by
> +.Xr packages-specs 5
5 -> 7
> +as being newer.
> +Once added, it cannot be removed or go backwards.
otherwise, ok jca@ for this hunk
> Defaults to empty (no need for numbering changes), then
> numbering starts at 0.
> Gets automatically incorporated into
> @@ -1658,6 +1662,12 @@ as
> .Sq v${EPOCH}
> to form a full package-name conforming to
> .Xr packages-specs 7 .
> +If there are dependent packages that use a version spec, this has to be
> +adjusted for each one or
> +.Ev PKGSPEC
> +changed on the affected package and
> +.Ev REVISION
> +bumped for all the dependencies.
Hmm, like Marc said this is not the place to mention PKGSPEC.
> .It Ev ERRORS
> List of errors found while parsing the port's Makefile.
> Display the errors before making any target, and if any error starts with
>
> [3. text/plain; packages-specs.diff]
> Index: packages-specs.7
> ===================================================================
> RCS file: /home/cvs/src/share/man/man7/packages-specs.7,v
> retrieving revision 1.25
> diff -u -p -r1.25 packages-specs.7
> --- packages-specs.7 27 Oct 2014 22:45:30 -0000 1.25
> +++ packages-specs.7 15 Mar 2020 11:25:27 -0000
> @@ -144,12 +144,46 @@ then normal version, and finally
> .El
> .El
> .Pp
> -In some rare cases, version numbering changes completely upstream.
> +In some rare cases, a change to a port would cause the version number to
> +compare as older than the previous version.
> +This happens if an update is reverted, if upstream's numbering scheme changes
> +completely, or if their usual scheme does not align with the one used by
> +.Ox .
ok jca@ for this hunk
> A version style marker, of the form
> .Sq v0 ,
> .Sq v1 ...
> can be appended to the version number (after the patch level)
> to denote the new numbering scheme.
> +This is added by setting
> +.Ev EPOCH
> +in the port Makefile.
> +Combined with
> +.Ev REVISION ,
> +this leaves the version ordering as follows:
> +.Bl -dash
> +.It
> +"0.1" is older than "0.1p0"
> +.It
> +"0.1p0" (or
> +.Sq p1 ,
> +.Sq p2 ... )
> +is older than "0.2"
> +.It
> +"0.2" is older than "0.1v0" (or any other version number without
> +.Sq v )
> +.It
> +"0.1v0" is older than "0.1p0v0".
> +Just as showed in the first point.
> +.It
> +"0.1p0v0" is older than "0.2v0".
> +The same as the second point.
> +.It
> +"0.2v0" is older than "0.1v1" (or any other version with a lower
> +.Sq v
> +value)
> +.El
> +.El
> +.Pp
I find this a bit too chatty. Can we find a single sentence that
explains what happens?
...
A version style marker, of the form ‘v0’, ‘v1 ...’ can be appended to the
version number (after the patch level) to denote the new numbering
scheme.
+This marker takes precedence over the regular package version.
Thoughts?
> See
> .Ev EPOCH
> in
>
> [4. text/plain; ports-guide.diff]
> Index: faq/ports/guide.html
> ===================================================================
> RCS file: /home/cvs/www/faq/ports/guide.html,v
> retrieving revision 1.89
> diff -u -p -r1.89 guide.html
> --- faq/ports/guide.html 28 Sep 2019 21:35:05 -0000 1.89
> +++ faq/ports/guide.html 14 Mar 2020 14:39:36 -0000
> @@ -1131,6 +1131,8 @@ changes to patches or build flags.
> If the upstream version has not changed, the package name bump is done
> by incrementing <code>REVISION</code> if already present, otherwise adding
> <code>REVISION = 0</code> towards the top of the Makefile.
> +This will add <code>p${REVISION}</code> to <code>FULLPKGNAME</code> to form
> a full
> +package-name conforming to packages-specs(7).
>
> <li>
> Version numbers always go forward.
> @@ -1138,6 +1140,8 @@ If something unexpected happens and you
> upstream numbering changes completely so that the version number appears
> to go backwards, you must use <code>EPOCH</code> to make sure pkg_add(1) sees
> the package as a newer package.
> +This will add <code>v${EPOCH}</code> to <code>FULLPKGNAME</code> to form a
> full
> +package-name conforming to packages-specs(7).
>
> <li>
> If the package does not build, no bump is needed: changes to restore a port
> @@ -1153,6 +1157,11 @@ The package system uses a signature mech
> is fully identified by its package names, plus the dependencies against
> which it was built, plus the version numbers of all shared libraries it
> contains.
I'd be tempted to leave this as is, or only add something like what you
propose below. I'm not objecting though. :)
> +
> +<li>
> +The version comparison ordering can be summarized as:
> +<code>0.1 < 0.1p0 < 0.2 < 0.1v0 < 0.1p0v0 < 0.2v0 <
> 0.1v1</code>.
> +Check packages-specs(7) for more detailed information.
> </ul>
>
> Part of the work will happen before the update itself.
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
