Digests send all posts

Confirming that mail-out.pl is in fact sending every mail everytime not just the new stuff since the last sending.

I confirmed a post from the 28th is being send every day. Moving to critical.

Kieran

What tables are necessary for a lightweight debugging environment? OG list manager takes a while to set-up and I'd like to quickly create an outbound mail only testing environment.

| og |
| og2list_groups |
| og2list_incoming_content |
| og2list_incoming_groups |
| og2list_msgid |
| og2list_outgoing_arguments |
| og2list_outgoing_content |
| og2list_outgoing_recipients |
| og2list_references |
| og_term |
| og_uid |
Node

Any other necessary tables?

The code applied to the digest block of mail-out.pl was seemingly borrowed from single mails and was expecting a single mid instead of an array of mids.

I've changed the way the mids are passed to the various send methods, and under the assumption that there will be multiple mids for a given recipient, have changed the way the array of mids are parsed within each of the send methods.

I tested using only sendmail as I've not the ability to test other send methods or single mails.

Feedback is, clearly, greatly appreciated as I'm breaking new ground and trying to have an immediate positive effect using a limited test suite, as well as minimal understanding of the data model.

I tested the latest patch, one I got via email and it is now working as expected.

We probably want to revisit the testing methods to ensure this patch scales. In particular we need to ensure that group subscribers have digest selected.

We are currently testing the digest patch for performance. We are going to test with 10 000 users and 100 mails in a group. 5000 of the group members will use digests.

Kieran

Fixed hardcoded DB.

Changes have been made to mail-out.pl to address the following:

a) Create the recipient list once per digest based on the unique recipients in the outgoing_recipients table for the nid
b) Do not rely on the combination of mid and recipient address as the means of flushing the outgoing_recipients table.
c) Clearing recipients is now done based on digests. Too, I've removed the logic where digest mids are cleared in conjunction to a 'send_' method.

d) Lastly, all changes have been applied to the do_digest block and will therefore can have no ill effect on legacy single mail behavior.

Could you detail all the testing steps, for each of the test cases? It sounded like there were some possibilities that users switching between direct mail and digests could lose mail.

Have you done any performance testing to make sure you algorithms scale?

people could get some mails twice, but not lose any. if they change theor mode to digest after the last digest was sent but already new mail was sent to the group they will have receieved that mail already as single mails and will get it again as part of their digest.

applied second patch.

Waiting for some test results from others before closing.