[lsb-discuss] Clarification of general LSB requirements

[Aaron Sowry]

On Thu, 2013-07-11 at 00:20 -0400, Jeff Licquia wrote:
> If we were to adopt the sysexits.h return values for sendmail in the
> LSB, it would be fairly easy to implement in lsb-invalid-mta.  And it
> sounds like there could be a use case for including those specifications
> if your software is expected to react differently to different sendmail
> error conditions.

For me, this would be a very welcome change. I wonder if there are other
places they could be used too?

> And I note that lsb-invalid-mta could use EX_SOFTWARE or EX_UNAVAILABLE,
> which don't seem to me to be lying. :-)

Myahh... grey area, but okay, I'll bite. I'd still argue that this is a
poor interpretation of the spec, even if it is strictly compliant.

According to [1], an LSB-compliant system must provide a crontab as
specified by [2]. In turn, [2] stipulates that:

"If standard output and standard error are not redirected by commands
executed from the crontab entry, any generated output or errors will be
mailed, via an implementation-dependent method, to the user."

Now I realise that this is a description of how cron should behave, not
sendmail, and that it is valid for sendmail to return an error to cron.
However, returning an error intentionally, every time, reminds be
painfully of [3]. I think that setting a precedent by allowing this
behaviour is detrimental to the LSB.


[1]
http://refspecs.linuxfoundation.org/LSB_4.1.0/LSB-Core-generic/LSB-Core-generic/crontab.html

[2] http://pubs.opengroup.org/onlinepubs/7908799/xcu/crontab.html
(I know this is an old version of the SUS, but it's all I could find)

[3] http://xkcd.com/1119/