diff options
Diffstat (limited to 'TODO')
| -rw-r--r-- | TODO | 70 |
1 files changed, 70 insertions, 0 deletions
@@ -0,0 +1,70 @@ +Release Critical: +* bind the broadcast listener to fixed local port (8765) (otherwise not firewallable) (matz) +* chmod TMPDIR/native, as otherwise set umask prevent it from being readable by the client + I'm wondering if it doesn't make sense to leave the daemon out of the game and let the + client package the environment - it could either cache it in /tmp and leave a note in + ~/.icecream or it packages it right into ~/.icecream and use e.g. the inode numbers of + the compilers to verify it's the right environment - think NFS users changing machines. + This would simplify some things and avoid the above bugs (and it would make it more + convenient for users of /opt/gcc-cvs/bin too) +* Improve Documentation (cschum) +* make the protocol version an uint32, not a hand-build array. +* let the client specify it was compiled with recent compiler (supporting -param). If so, + let the daemon compile with the options, otherwise leave them out. + +Akademy must have's: +* the scheduler with the newest protocol version wins +* only one scheduler within the same broadcast domain +* daemon: only accept connections from a sane net prefix (scheduler can determine + this and send it out via confMsg) to avoid being exposed to the internet +* always re-discover scheduler after disconnect, don't just re-connect to the last + one (for some reason I don't see why this happened, but it did) +* benchmark/testing jobs (broken machines with broken ram and broken network card on the + icecream network destroy the fun and are not easy to find manually) + +Random: +* Option -iiface to specify the interface[s] to listen on for the scheduler, or + to use for the daemon. +* Don't explicitly forbid tunnels because it could be useful for things like + vmware +* If someone calls a amd64 client on a host that runs a ia32 daemon and there are + no other amd64 daemons in the farm, he will get no answer, but a timeout from + scheduler (quite a corner case, but neat) +* use syslog +* Log problems found at some scheduler log - or even in the monitor. E.g. if a client + can't reach a given daemon, it should be able to tell. Perhaps the scheduler can even + disable that very host for some penalty time +* Reduce amount of force-waits, especially if they involve network latency (scheduler queries) + and daemon context switches: + - remove the need for EndMsg + - do not ask the daemon about the first job (WIP dirk) +* if a compile job SIGSEGV's or SIGABORTs, make sure to recompile locally because it could + be just a glibc/kernel incompatibility on the remote site +* Add heuristic which optimises for overhead-reduction. Statistics prove ( ;) ), that for + e.g. linux kernel the file size varies a lot, and small jobs should be preferably compiled + locally and bigger ones preferably remote. +* Split number of jobs into number of compile jobs (cheap) and number of non compile jobs (can + be expensive, e.g. ld or meinproc). The reason is that multicore chips become more and more + common. Today you can get quad cores easily and in some month we've 8 cores and several link + jobs at the same time can pull down your whole system. + (I can life with a hard coded number of one non compile job but others probably prefer a + configurable number) + +Suggestions from "Wilson Snyder" sicortex.com: +- Add ICECREAM_RUN_ICECCD to make the scheduler machine not run iceccd +- Have schedulers redudant - e.g. as in nmbd + +Christoph Thielecke writes: +> 1) Problem von icecream: icecc hat /usr/bin/gcc-4.2 kopiert (ist aber ein +> Symlink auf den Hardening-Wrapper, der installiert war, siehe Listing +> unten). Kannst Du vieleicht eine Prüfung einbauen, ob gcc-x.y mit .real +> endet bzw der Wrapper ist? + +Andi Kleen writes to protect against misconfiguration: +>generate somefile.h on the client with +>#if __GNUC__ != expected_major || __GNU_MINOR__ != expected_minor +>#warning blabla +>#endif +>and then run with +>-include somefile.h +> (this file needs to be in the environment created by create_env) |