Make sure other tasks are processed with tasklists that contain both bad xfrs (that cause reload to exit) and other tasks

[wtoorop]

No description provided.

[wcawijngaards]

The code looks okay. I worry that the tasks processed by the old-main process take a long time, and that old-main is not responsive during that time. For tasks for checkzonefiles or writezonefiles, that also seems reasonable.

[wtoorop]

The code looks okay. I worry that the tasks processed by the old-main process take a long time, and that old-main is not responsive during that time. For tasks for checkzonefiles or writezonefiles, that also seems reasonable.

Thanks @wcawijngaards commit 0c4e275 addresses this issue you raised of notifies not being passed when non transfer tasks are being processed (and it works), although it's hard to write a unit test for this (but I've tested manually with a strategically placed sleep). If you can review again and approve (perhaps only the last commit), then I can merge this as solution for the netnod case.

[wcawijngaards]

Okay that looks nice and the server stays responsive to notifies. It creates another process, but it seems to also solve the responsiveness problem.

[k0ekk0ek]

I personally dislike the additional fork. IMHO it would be better to remove the NOTIFY proxy code from the main process and just let the serve child talk directly to xfrd, though there is probably a reason why it is the way it is too.

[wtoorop]

I personally dislike the additional fork. IMHO it would be better to remove the NOTIFY proxy code from the main process and just let the serve child talk directly to xfrd, though there is probably a reason why it is the way it is too.

Passing NOTIFY from serve childs to xfrd is developed in branch devel/stream_notify2xfrd.
PR #390 is this PR rolled back to commit 22addb5 and with the devel/stream_notify2xfrd merged in. It replaces this PR.