Date: Wed, 12 Jun 91 09:14:03 GMT Message-Id: <2145@n1bee.ampr.org> From: n1bee@n1bee.ampr.org (Mike Bilow - Cranston, RI - Sysop @ [44.104.0.20]) Reply-To: n1bee@n1bee.ampr.org To: ka1xn@ka1xn.ampr.org, wc1r@wc1r.ampr.org, n1hid@n1hid-1.ampr.org, w1imm@w1imm-2.ampr.org, kz1f@kz1f.ampr.org, wa1equ@wa1equ.ampr.org, w1cg@w1cg.ampr.org, ka1jy@ka1jy.ampr.org, ka1mf@ka1mf.ampr.org, n1bee@n1bee.ampr.org Subject: GRINOS v0.2s released I am going to start an FTP upload of the new GRINOS-S.ZIP to WB6NIL as soon as I am done typing this message. This version has significant enhancements, including: -- A major bug fix caused by an incompatibility between the old and new Borland C compilers, in collaboration with NOS replacing a library function or two. This caused TMP##.$$$ files to be created but not deleted. This bug has been repaired, but at the cost of no longer supporting the user option to explicitly specify the directory to be used for temporary files. All temporary files will be created in the current directory. -- Correction of a stack framing error also related to the use of the new Borland C compiler. This may have been responsible for the problems seen with the garbage collection routines for NOS memory. I am still not confident of the garbage collection code. -- Auto detection of machine class. The "isat" flag should now be set correctly for the type of processor found during initialization. As I explained earlier, it is now required that the "isat" flag be set correctly for proper hardware interrupt handling, and this will relieve most users of the need to worry about it. The "isat" flag can still be changed manually with the "isat 1" or "isat 0" commands for cases where the auto detection fails for some reason. -- Memory efficient mode now defaults to enabled. I have already recommended the use of efficient mode, although it can still be changed by command. -- Enhanced reporting of information with the "memory status" command. Efficient and circular modes now show as enabled or disabled, and the current memory threshold is displayed. Remaining known bugs: -- I still have not fixed the problem with shelling out to DOS or the mailer in the first ten minutes or so before NOS has stopped calling for more core. -- Memory garbage collection may still be broken. Immediate future plans: -- A "tcp blimit" command that would provide a maximum backoff. There has been substantial demand for this feature. This version is designated KA9Q 910531/PA0GRI v1.6h/N1BEE 910612v0.2s. This is a beta-test release. That means it should not be used on unattended switches. -- Mike