... forgot to CC the list and other recipients, sorry ...


[... Apache::Test test failure ...]
Why is it so hard to make installation scripts work unattended?


IMHO, you're asking the wrong question ... Why should the
onus be on the author of the module to handle cases like
this? Couldn't the automated testing script be adapted to
recognize that such a loop has been entered, and infer from
that that some human intevention is required to give some
sensible information, and then abort? Apache-Test isn't
unique in this respect. Asking to insert extra logic in a
package that would be used solely to satisfy a situation
that no typical user would encounter isn't reasonable.

This is not only about the automated testing (which wouldn't happen in
my case anyway, as there is nothing resembling Apache installed on my
Win32 box) but also about unattended install - the Makefile.PL can't be
run without a user in front of a terminal, which I consider bad, and not
really atypical, but that might be my atypical situation of not sitting
in front of a connected terminal when a computer does things.

In my opinion, a script should ask questions with sensible defaults and
never delve into an endless loop, asking the same questions again and
again. Either quit on the first time around, or detect that there is no
user to communicate with anyway and bail out if the sensible defaults
don't work. Which is additional work, yes, but can be expected IMO.

-max



Reply via email to