On Sunday 20 August 2006 11:48, David Härdeman wrote: > On a more general note: is this something that you'd like to see in d-i > before the Etch release? Yes, it is. It will help to reduce the minimum required memory a bit. (Unless Christian keeps adding translations of course ;-) > I think I'll add a naive check of the "MemFree" field in /proc/meminfo > for now. If the memory seems low, it can warn the user that things may > crash and let him choose to allow the installation to continue or to > not use crypto. OK. > >Also, is there a possibility that the function will be called again > > later? In that case it could be good to prevent running the memory > > check and loading the modules a second time. > > Ok, I'll add a check for that. I thought that anna wouldn't re-install > the udeb if called a second time. No it probably won't. However, executing the memory check a second time could result in an error even though the modules have already been loaded. It seems cleaner to me to have the test anyway. > >Finally, wasn't there some code that checked if everything was > > available before offering a method? Does that still work with this > > change? > > The previous suggestion tried to load all methods as soon as crypto was > chosen for any partition. This leads to additional downloads (since you > will probably only want to use one method), memory usage and waiting. The reason I asked is that anna will probably not complain if dependencies are not available and thus I think you have to check if they really were loaded some other way. I'm not completely sure about this though, but suggest you keep it in mind.
Attachment:
pgp4oj8xatLCH.pgp
Description: PGP signature