Pre-Elan second message occasionally stomps over Pre-Elan first message

[SamStudio8]

The #inbound-distribution message sometimes announces an empty site list as the file has been modified by the later running second message. This is usually during times of heavy I/O load reducing the efficiency of find (see #19). The two message steps should use a different working copy, or the second message should be held at a signal until safe to proceed.

[SamStudio8]

Added to backlog #62

[SamStudio8]

I have been sentimental about this bug because the occasional announcement of 1 and not much else is pretty amusing. However I've just noticed that this week week the full debug messages to #inbound-distribution have not been appearing anymore. My suspicion is the stomping window has aligned such that the LONG message is completely suppressed by the SHORT message.

[SamStudio8]

Naturally nobody has noticed this, but it ought to be fixed giving it's causing real trouble rather than whimsy. I'm taking the opportunity to merge the two different messaging routes (caused when Elan was put in charge of sending the morning SHORTSTART announcement).

[SamStudio8]

SamStudio8/elan-nextflow@d20cf00 removes the messaging code from go_count_and_announce and uses the standard message_uploads script. Additionally, the temporary files are stamped with the message type to prevent interaction.