Package number arithmetic

[naddy@mips.inka.de (Christian Weisgerber)]

Experimentally, I did a complete package build run on i386.  Some
figures:

a) INDEX:                       1427 packages
b) actually built:              1297 packages
c) in INDEX, but not built:       46 packages
d) logs:                        1324 packages
e) broken, from build log:        16 packages

If you can't make those numbers add up, you're not alone. ;-)

"sort -u INDEX" turns out only 1345 packages.  MULTI_PACKAGES create
duplicate index entries.

Checking INDEX with duplicate lines removed against packages/i386/All
yields 1299 successfully built packages.  Wait, we have only 1297
packages.  Here's the offenders:

python-tools-2.0        lang/python,-tool
python-tools-2.0        lang/python,-tools,threads
python-tools-1.5.2      lang/python15,-tools
python-tools-1.5.2      lang/python15,-tools,threads

Somehow the thread flavor doesn't show up in the package name.  Bad.

Out of 1343 packages, 29 don't show up in individual portslogger
log files.  Those are from ports that are marked BROKEN or for
other architectures, so they never emit the "===> Extracting" marker
portslogger uses.

1314 packages create 1324 log files?

There are five bogus log files:
for erlang-47.4.0.log
for exmh-2.0.2.log
for nscache-0.2p1.log
for pkfonts300-1.0.log
for wine-990225.log

(Yes, "for erlang-47.4.0.log" etc.)

security/nessus triggers the creation of bogus separate log files:
libnasl.log
nessues-core.log
nessus-libraries.log
nessus-plugins.log

So does net gaim:
pixmaps.log

1314 log files minus 16 build failures document the building of
1297 packages.  Hmm.

There it is, mail/exmh2 has NO_PACKAGE set.

All accounted for.

-- 
Christian "naddy" Weisgerber                          naddy@mips.inka.de