jolley@phcoms.seri.philips.nl (Graham Jolley) (12/12/90)
I am having a problem building minix-386 with more than about 55 buffers. I remember a posting that said this was due to a problem with the 16bit build silently clobbering one or more 32bit values. The thing is I'm getting this problem even though I am now using a 32bit build ie I am building a new kernel while running minix-386. I am using the 'normal' build procedure NOT shoelace. I notice that one of the bss sections is at 64K when the problem occurs. Is this significant? I apologise in advance if this question has already been asked and replied to. Does anybody have a copy of the 'db' code they could send me - my 'db' seems to be a dummy? Also have there been any new versions of bcc or the 386-kernel made available, my versions are of uncertain origin. The warning messages from the compiler about redefinitions are a bit of a pain :-). Finally like everybody else I think the Bruce Evans '386 extensions and bcc are very nice pieces of work. Thanks. Graham Jolley
wkt@ccadfa.adfa.oz.au (Warren Toomey) (12/14/90)
In atricle <565@phcoms.seri.philips.nl>, (Graham Jolley) writes: > I am having a problem building minix-386 with more than > about 55 buffers. > I am using the 'normal' build procedure NOT shoelace. From memory (and Bruce will definitely know), you must make a 32-bit version of build, and then build an image using it. Something in build stops it from working as a 16-bit binary. The procedure: Under 16-bit mode, make the 32-bit kernel with 30 buffers Boot the 32-bit kernel Compile build as a 32-bit binary Recompile the kernel using 300 :-) buffers Using the new build, build a new image (keep the old one!) Boot the new 32-bit kernel. I have a 32-bit kernel using this method (and no Shoelace as yet), and it works fine! -- Warren Toomey VK1XWT, about to Appdiff Deep in the bowels of ADFA Comp Science. `TCP/IP? Really?!'