Module site offline indefinitely

[Jon South]

Actually, I think you slightly misunderstand what I'm suggesting-- not a
new tool or review process; our scripts should fail on these while a person
is testing their changes, long before they try to commit.

These problems actually output error messages, but they are not being
caught is because lunar/theedge is not capturing them or receiving a bad
return status. This is caused by the same issue when BUILD is created
without using && for command chains, so when a typo exists in the middle,
the error status is gobbled by the next line. In testing, using && causes
lin/lvu/etc to properly fail when DETAILS is read. The problem here is not
that unique, except I don't think we want to start forcing && at the end of
every line to resolve it (though it would permanently fix it) as it would
require tedious changes and appear rather ugly.

My suggestion would be this: the core tools need to be checking for output
on stderr when sourcing DETAILS, as there should *never* be stderr output
when doing so. This will help people making or updating modules to find
these problems during testing. I believe this would be the most practical
and easy way to go about it.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://foo-projects.org/pipermail/lunar-dev/attachments/20130909/596c5f95/attachment.html>