Fails saving Project

A

aworonow

Well-known member
Win 10, latest updates.
PI 1.8.8-5

PI cannot save a project to a previous or new file/location because a .swp file cannot be found. the swap file was on ram disk, and the ram disk closed...hence the swap file was gone.

A. is there a work-around / recovery I can used to keep the full project at this point?
B. is there a fix that could be implement in PI to circumvent this error occurring?

Thanks, Alex
 
the swap files *are* the project... if they are gone there's nothing to save.

i have been preaching this up and down, swap performance is just not that important relative to processor performance. imo it's a bad idea to keep these files on a volatile medium for the tiny boost in speed they represent.

generally speaking an operating system won't let a file be deleted if it's opened by a process... but i suppose there's nothing stopping either a bug from causing the ramdisk to unmount spontaneously or the user forcing the disk to be unmounted.

are all the swap files gone? meaning were some being kept on a non-ramdisk? from the name of the missing swapfile you can probably infer the view it belongs to. if you close that view (and any other view that had it's swap files go away) then you may be able to save the project. but it will be an iterative process, closing views and then trying to save the project again and inspecting the error message to figure out the next view to close. and of course you've lost all the undo history for all of those views.

before closing a view you can clone it so at least you'll have the last state around when you are finally able to save the project. hopefully PI won't try to create the swap file for the clone on the missing ramdisk.
 
Thanks. I've learned a lesson. I thought I had read that swapping was the most critical process in determining speed.
Fortunately, I kept individual, major change frames along the way (to a hard disk), and could always save some of the critical icons too.
 
it boosts the benchmark numbers a lot but personally i don't think that reflects reality. per process invocation PI at most does one read and one write from/to the swap directory. if your machine spends 10s on deconvolution and then 0.5s writing/reading swap, that's not a whole lot different if it spends 10s and then 0.1s as far as the user experience goes. also there's lots of "think time" involved in processing so that's likely to contribute quite a bit as well.
 
You must log in or register to reply here.
Back
Top