Saturday December 11, 2010

[Time] NameMessage
[11:43] mikko sustrik: you there?
[11:43] sustrik hi
[11:44] mikko i created a small patch for the swap
[11:44] mikko now it fails when you try to set invalid swap file. the implementation doesn't feel ideal but it works
[11:44] sustrik mikko: i'll review it
[11:44] sustrik just send it to the ml
[11:46] mikko is there a portable PATH_MAX ?
[11:57] sustrik mikko: let me see
[11:58] sustrik it's POSIX
[11:58] sustrik
[12:18] mikko it's there for now. does it look sensible?
[12:19] mikko i'll mail ml later, now i need to vacuum before my better half comes back from hair dresser :)
[12:31] mikko hmm
[12:31] mikko maybe it would make more sense to move swap to options
[12:31] mikko and initialize it when user sets swap size
[12:34] sustrik mikko: yes, that's what i had in mind originally
[12:34] sustrik there's one problem though
[12:34] sustrik on the bind side there are N connections
[12:35] sustrik the number is unspecified
[12:35] sustrik so you cannot create the files in advance
[12:35] sustrik maybe creating a single dummy file to check the permissions would do?
[12:37] mikko that should be fine
[12:37] mikko or maybe even check if the folder is writable
[12:38] mikko not sure how that works with windows acls etc
[12:38] sustrik if possible with POSIX API...
[12:38] sustrik creating a dummy seems easier imo
[12:40] mikko should the swap append sequence number to filename given?
[12:41] mikko so if i give /tmp/test.swap the actual filename would be /tmp/test.swap.1
[12:41] mikko or something
[14:46] mikko sustrik: i think swap_dir option makes more sense than swap file