I run a debian derivative called raspbian aimed primerally at the raspberry pi.
While building ngircd on the raspbian jessie autobuilders I got a testsuite failure with an explicit request to report to this email address, so I am doing so.
running misc-test ... failure! FAIL: misc-test <--snip--> running who-test ... failure! FAIL: who-test <--snip--> =========================================== 2 of 16 tests failed Please report to ngircd-ml@ngircd.barton.de ===========================================
The full log is available at http://buildd.raspbian.org/status/fetch.php?pkg=ngircd&arch=armhf&ve...
peter green wrote...
While building ngircd on the raspbian jessie autobuilders I got a testsuite failure with an explicit request to report to this email address, so I am doing so.
running misc-test ... failure! FAIL: misc-test <--snip--> running who-test ... failure! FAIL: who-test
<--snip-->
2 of 16 tests failed Please report to ngircd-ml@ngircd.barton.de ===========================================
Thanks for the feedback. I was able to reproduce the issue on my pi in a wheezy schroot.
That's interesting since building on Debian armel and armhf went through without problems[1]. Requires more checking, will do. Stay tuned.
Christoph
[1] https://buildd.debian.org/status/package.php?p=ngircd&suite=sid Yes, I known Debian's definition of "armhf" is different.
Christoph Biedl wrote...
That's interesting since building on Debian armel and armhf went through without problems[1]. Requires more checking, will do. Stay tuned.
... aaaaaaand as I feared: The ngircd test suite relies on certain settings in /etc/hosts. If they are not in the way expected, some tests fail. That's why I removed whois-test a long time ago in the Debianization. These two appearently fail on
| 127.0.0.1 raspberrypi localhost
in /etc/hosts. Dirty workaround: Remove that file before building. Worked for me.
test log for misc-test:
(...) users :ngircd.test.server 446 nick :USERS has been disabled userhost :ngircd.test.server 461 nick userhost :Syntax error userhost nick :ngircd.test.server 302 nick :nick=+~user@raspberrypi
The last reply does not match
-re ":ngircd.test.server 302 nick :?nick=+.*@(localhos.*|127.0.0.1)"
test log for who-test:
(...) who #channel :ngircd.test.server 352 nick #channel ~user raspberrypi ngircd.test.server nick G@ :0 Real Name :ngircd.test.server 315 nick #channel :End of WHO list mode #channel +v nick :nick!~user@raspberrypi MODE #channel +v nick who #channel :ngircd.test.server 352 nick #channel ~user raspberrypi ngircd.test.server nick G@ :0 Real Name :ngircd.test.server 315 nick #channel :End of WHO list who localhos* :ngircd.test.server 315 nick localhos* :End of WHO list
The last reply does not match
":ngircd.test.server 352 nick * * * ngircd.test.server nick G :0 Real Name"
Alex, I think the test suite needs more flexibility. There are several ways how /etc/hosts might look and AFAIK there's no rule or even consensus how it's done right. Basically it's a question of whether the first line resolving 127.0.0.1 contains "localhost" or a host name.
Instead of creating rules for the world here it seems more sensible to either relax all checks to not strictly check for localhost, or do a lookup first and patch the expect scripts accordingly.
Just out of curiousity, what actually prevents the tests from using IPv6 if available, and breaking horribly then?
Christoph --
-
Christoph Biedl -
peter green