Development server kills itself to reload, causing browser GET failures

[aaronsw]

The current development server seems to use a threaded model that kills and restarts the server thread everytime the code changes. Since it takes some time to startup again, this leaves some downtime between when you hit save and when you can view the page. While such a delay is unavoidable, with this model it manifests itself as an inaccessible page as opposed to (if, e.g. CGI was used) simply a delay. This kind of annoyance is simply deadly for development, since it discourages making small changes.

[aaronsw]

This is driving me up the wall.

[eugene@…]

It works for me. Reload time is instanteneous. I never had problem with GET failures. Probably you have some CPU-intensive code running during reloading.

[sune.kirkeby@…]

If you are on a UNIX-like platform you can try ibofobi.utils.developer_http,
it never closes the socket when reexec'ing, so it should never miss a connection.
You can use that, until Djangos developer HTTP is fixed.

See ​http://code.ibofobi.dk/public/file/trunk/ibofobi/src/ibofobi/utils/developer_http.py.