Testing for the Linux Standard Base 3.0 [LWN.net]

Testing for the Linux Standard Base 3.0

Posted Sep 20, 2005 5:35 UTC (Tue) by ajosey (subscriber, #1938)
Parent article: Linux Standard Base 3.0 released

Ulrich is right in that more investment in test development is needed. A number of tests were contributed about 4 or 5 years ago but have not had an update to properly track the standards, keep up with hardware developments or allow for new implementations that are conforming yet different. Also many of the tests were written as source tests for API testing, and hence when compiled into a binary can get timing problems, there have been some attempts to address this but Moore's Law seems to win at times.

(Log in to post comments)

Testing for the Linux Standard Base 3.0

Posted Sep 20, 2005 6:13 UTC (Tue) by khim (subscriber, #9252) [Link]

If test does not work on today's 64-way NUMA SMP system then test is incorrect - and it does not matter if it worked before or not. Most tests are not even checking correct things: they test simple things declared by specifications and often are doing it incorrectly. What they do not test is obscure bugs in different parts of GlibC and when we are talking about GNU/Linux systems where 99% of systems are using one form of GlibC or another it's more important: if LSB-environment is not the same as used not "native distribution programs" you are essentionally make LSB-compliant programs second-class citiziens with not support. Hardly a way to go...

I can not agree with Ulrich 100% - may be LSB can be usefull for something but right now... I found that my Gentoo system (Gentoo refuses to spend time and resources on things like LSB; FHS - yes, LSB - no) works with third-party programs just fine and often better then some LSB-certified distributions.

Testing for the Linux Standard Base 3.0

Posted Sep 20, 2005 16:40 UTC (Tue) by smitty_one_each (subscriber, #28989) [Link]

I think the question of whether the test suite at hand is incorrect is a separate issue from whether or not LSB is a good idea.
If you want to fuel a Redmond argument that a license from Linux vendor L breeds equally bad lock-in as a proprietary solution, then, by all means, blow off LSB.

Testing for the Linux Standard Base 3.0

Posted Sep 21, 2005 17:25 UTC (Wed) by khim (subscriber, #9252) [Link]

I think the question of whether the test suite at hand is incorrect is a separate issue from whether or not LSB is a good idea.

Suuure. Does it mean you are now using OSI networking ? I like IETF approach better: if there are no working code then it's irrelevant if core idea is good or bad. Once we'll have working code we can talk about it. This approach actually works. And LSB does not have working code right now (the thing "sorta works" like X.400 but mostly it's just pretty picture you can print on package of your distribution) - thus it's useless to discuss "the idea".

If you want to fuel a Redmond argument that a license from Linux vendor L breeds equally bad lock-in as a proprietary solution, then, by all means, blow off LSB.

Huh ? What is this all about ? Do we need to call black white just to avoid Redmond's criticism ? This is strange notion to me. LSB does not work. Period. If you do not like it - then fell free to offer solution, may be in the end it'll be usefull. But as long as LSB certificate is just pretty picture it's just waste of resources.