[HEADSUP] portsmon default ports environment switched to i386-7
linimon at hub.freebsd.org
Thu Oct 11 22:33:47 PDT 2007
(For those who are not familiar with the FreeBSD Ports Monitoring System,
or portsmon, that I wrote and maintain, you'll find more information at
To get ready for the 7.0 release, I've switched its model of the "default"
ports build environment from i386-6 to i386-7. This mainly affects the
status metavariables of each port. (It's a current limitation that portsmon
only models one build environment.) E.g., its reports of ports that are
marked "BROKEN" (i.e. fail to build for some reason) is now based on the
evaluation of their state on -7 rather than on -6. This applies to both
the interactive pages, and the email that is sent out every 2 weeks to
maintainers and the mailing lists.
Some of our maintainers will be seeing these messages for the first time.
Don't panic; in many cases the newly-marked ports are failing to build
because of the import of gcc4.2 into the base system. (All of the ports
that are failing on i386 and/or amd64 have already been marked). We still
have time before the 7.0 release to fix these problems -- most have already
been fixed over the past few months.
For some outdated software, it may not be feasible to provide patches; if
this is the case, we can specify that the port can only be built with an
older version of gcc. However, this should be used as a last resort. But
the preferable solution is to provide patches and then send them to the
upstream maintainers (if any) so that they can be incorporated there.
In addition to the pages that most people are familiar with, I am in the
alpha stage of generating reports that show the state of ports across all
build environments. Unfortunately, the results are not yet dynamically
generated. I had hoped to have that ready before release. In the meantime,
the static reports will have to do.
compares the state of the packages for each build environment. The
bars in the chart are divided up into:
- unrestricted packages built;
- restricted packages (built, but not available for download);
- build errors;
- not tried due to BROKEN, IGNORE, or FORBIDDEN;
- not tried due to NO_PACKAGE;
- not tried, other (these are due to drift in the ports tree between
the time it was checked out, and the time it finished);
- not tried, because it was a dependent port of one of the 3 lines above;
- missing. These are due to bugs in the algorithm.
- Clicking on each bar in the above will take you to a page for a particular
build environment with a pie chart for the above values, and a tabular
representation. (Bug: the links on those pages do not go anywhere
- You can also (for the first time) see a tabular list of all ports that
fail to package -- for the union of all the reasons above. Examples:
Here you can see, for a given build environment, the packages that were
not built; the reasons that they were not built; and all the ports that
depend on them. Clicking on each link in the left column will take you
to the overview page for that port.
Fair warning: these charts are pretty large; you may need to shrink them
down to browse them effectively.
Further note: some of the ones I have just uploaded show the results of
a few very important packages that failed (x11/kdelibs3; devel/imake;
and graphics/ImageMagick), so the statistics look bad. However, I
believe these ports have already been fixed.
Again, these reports are in an alpha state, but they offer a way for people
interested in seeing e.g. what the state of the amd64 packages will be for
7.0, and hopefully give them a chance to fix some of them before the release.
Finally, you don't need to let me know about the specific bugs in these
reports -- I'm well aware of them :-)
More information about the freebsd-ports