A reminder to "FBBG" (sometimes!) and a question about the eudev
testsuite.
On my recent builts, I've been getting a couple of udevd error
messages during the boot, of the form
udevd[1556]: IMPORT{builtin}: 'keyboard' unknown
/lib/udev/rules.d/60-keyboard.rules
and referring to lines 16 and 20 of that file.
Initially, I put that down to my variations (omit what I can from
the packages added when we trying to do dual systemd/sysvinit, and
omit things which do not give me any value.) So, I was not building
keymap within eudev unless I did a straight "by the book" build - I
had assumed that I could drop gperf and intltool from server builds,
although in practice gperf and intltool are used for my desktop
builds.
Then last weekend I realised that one of my May builds was indeed
"by the book", but it still had this message (with eudev-1.6). I've
been searching around, and basically got no results beyond noting
that /lib/udev/keymaps no longer exists, even though the eudev
configure reports it will be used (and even for --disable-keymap!),
In the end, on my current builds (eudev-1.9, still managing to omit
acl) I have rebuilt eudev without --disable-keymap : when shutting
down I got similar messages about 'keyboard' unknown (not noticed
those before on shutdown), but the next boot did not have those
messy error messages.
For the May build "by the book" with eudev-1.6 or whatever it was,
I guess that upgrading eudev would have sorted it.
So, I have decided to always build gperf and intltool, even for a
server.
This prompts my question about eudev's testsuite : when
reinstalling eudev as in chapter 6, the test results were fine:
make[4]: Leaving directory '/building/eudev-1.9/test'
============================================================================
Testsuite summary for eudev 1.9
============================================================================
# TOTAL: 2
# PASS: 2
# SKIP: 0
# XFAIL: 0
# FAIL: 0
# XPASS: 0
# ERROR: 0
============================================================================
instead of the normal PASS 1, FAIL 1 result.
That encouraged me to consider what might have changed (the system
is built and booted). But then I realised that my rebuild had not
enabled gudev, although various desktop packages want it. So, I
decided to rebuild again, without disabling keymap, but I am back to
a more normal
============================================================================
Testsuite summary for eudev 1.9
============================================================================
# TOTAL: 2
# PASS: 1
# SKIP: 0
# XFAIL: 0
# FAIL: 1
# XPASS: 0
# ERROR: 0
============================================================================
See test/test-suite.log
Please report to https://github.com/gentoo/eudev/issues
============================================================================
Makefile:850: recipe for target 'test-suite.log' failed
make[3]: *** [test-suite.log] Error 1
and so forth.
I also recall that I think Armin said systemd has a similar
"failure".
I did save test/test-suite.log each time, but apart from sporadic
messages of the form
open /dev/null failed: No such file or directory
and various error messages for tests 78 to 82 which also happen in
the "good" run from LFS chapter 6, all I can find is the final
1 errors occurred
Should I just write these test results off as "stuff happens" (or
"occasionally, stuff works if you are lucky") ? ISTR that one of the
eudev tests is reputed to use Python - which is obviously not in
chapter 6 - but that doesn't explain why the tests from the rebuild
with gudev enabled fail both times.
ĸen
--
Nanny Ogg usually went to bed early. After all, she was an old lady.
Sometimes she went to bed as early as 6 a.m.
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
