First, i have to say that situation with postcards has improved considerably
lately
Thanks to everyone who I did not find time to reply personally.
Now to the business. Nes version 1.02 is available for download.
http://members.home.com/kovalenko/software/spin-1.02.tgz
It has about dozen improvements and fixes, some of them important enough.
Here is full list of changes:
- Added accounting (rather kludgy) of memory consumed by process manager.
- Implemented proper resize handling. SPIN does not restart on SIGWINCH
anymore. - Resize feature will be autodisabled on terminals which don’t support
TIOCGWINSZ. - Added options for rows & cols. Cols is ignored but either option disables
resize handling. - Machine architecture now reported along with node name.
- Removed check for root ID. If non-root swap stats won’t be displayed.
- Implemented unified normalization for K/M/G/T-byte values.
- Failing MsgSend to procnto is not treated as fatal anymore. It fails on
anything but x86. - Added kludge to report real procnto name (procnto-smp, procnto-600, etc).
- Added workaround for sprintf() problems on PowerPC (endianness I guess).
- Added support for PowerPC in makefile. SPIN has been tested on MCP750/765
with NTO2.11 - Added PowerPC binary.
So, SPIN is now ‘officially’ multiplatform. Aside from PowerPC it is
reported to work on SH as well. Few things to consider:
-
Proper handling of resize has downside of being subject to ncurses bugs.
Some terminals works better, some worse. I find that vt100 and xterm are
best TERM types to use. QANSI has problems. -
The code data and heap size reported for procnto now are somewhat
‘artificial’. They are not reported properly by procfs and won’t be until
new VM is in place. Thus, there’s kludge to report something meaningful.
Here is convention:
- code size is what 'ls -l ‘/proc/boot/procntoXXX’ reports.
- data size is equivalent of ‘df /proc/boot’ - code size.
- heap size is equivalent of ‘df /dev/shmem’.
This gives fairly close estimation and allows you to track memory better.
Previously you’d not see memory consumed by procnto at all, so if you copy a
big file to /dev/shmem you’d not see any changes in SPIN output. Now you
will.
Have fun,
- Igor