home *** CD-ROM | disk | FTP | other *** search
- (30 September 1991)
-
- July 1991
- simple printf server that starts own client (done)
- design changes to dev and net modules (dev: need to understand how
- to do tty's; net: need to understand???) (done)
-
- August
- client can map file read/write into its address space (done)
- VM error handling (permissions, non-existent pages) (done)
- server has no-op locking calls in place (done)
- client is backed entirely by single-threaded Sprite pager (done)
-
- September
- ds5000 running Mach 3.0 installed, but not used for Sprite work
- (not done; got Mach 3.0 running on Ultrix disk, but need to get
- second disk to fully install Mach)
- server can read time-of-day chip (not started)
- server is multi-threaded, with coarse-grained locking (done)
- client can fork and exec children (fork done, but needs more
- testing; exec needs small amt. of server changes and needs user
- code)
-
- (Also done)
- trivial exception handler (server code to handle user address
- faults, etc.)
- small libc started, with all knowledge about Mach and RPCs moved
- into libc (from October list)
- Proc_ServerProcs working to "immediately" start new process
- (time-based calls aren't implemented) (head start on November)
- fixed gdb (under Mach) to print the thread names registered by the
- Sprite server
-
- (unexpected problems)
- Setting up cross-development environment (on Sprite) for user
- programs. In particular, getting lint support and cleanly
- separating the Sprite environment from the Mach environment (e.g.,
- stdlib.h versus libc.h) were both a pain.
- Ultrix 4.0 incompatibilities with Mach
- couldn't reliably get back to gdb command prompt to find out what
- Sprite server was doing (fixed thanks to patches from OSF
- Grenoble).
- stress testing of Sprite server tends to crash Unix server
- covering potential race conditions (segment management and process
- management) was harder than expected.
-
- (net estimate)
- About a week's worth of stuff hasn't been done yet that was
- scheduled to be done.
- Around 3-4 days' worth of stuff has been done that was scheduled to
- be done later.
- Therefore, I'm a couple days behind schedule.
- --------------------
- October
- doing all work on ds5000
- server can ping a native Sprite system (i.e., RPC's work)
- client makes system calls through emulation library, which then
- invokes MIG stub
- (SOSP)
-
- November
- Proc_ServerProc's work (change timer queue to use message w/ timeout)
- file system works (client can read & write file from Sprite file
- server; cache memory is static & pinned)
-
- December
- client can page across network
- start on signals
- (short Christmas vacation)
-
- January 1992
- client can send, catch signals
- (XPRS retreat)
-
- February
- work on C library
- access to console device.
- be able to run gcc, maybe vi.
-
- March
- run benchmarks, find bottlenecks
- start fixing bottlenecks
-
- April
- more benchmarks, performance work
-
-
- Dropped:
- access to local disk
- binary compatibility
- support for X
- TCP/IP support
-