More findings about SMSQ/E v3.01 for SGC: The sources, as supplied by Wolfgang won't link properly ! The "smsq_gold_roms" routine is not linked by "flp_bas"... I added the missing line and I now got a SMSQ/E version which _tries_ to initialize the ROMs on my SGC/Aurora/Qubide/SH/RomDisq system: no joy though...
When the IDE cable is plugged, I get the same behaviour as what described in my previous message. If I unplug the IDE cable or RomDisq, SMSQ/E boots and tries to initialize the ROMs, but I just get a blank screen and it crashes even before the first ROM copyright message is displayed... Same thing if I try to boot SMSQ/E v3.01 from a working 2.91 session... I'm puzzled ! I wonder if it's something dealing with caches (were the caches disabled during ROMs initialization in SMSQ/E versions prior to v2.98 ?). Thierry. ---------------------------------------------------------------- Below is my previous message for reference. ---------------------------------------------------------------- Begin forwarded message: Date: Fri, 8 Aug 2003 04:00:18 +0200 From: Thierry Godefroy <[EMAIL PROTECTED]> To: ql-users <[EMAIL PROTECTED]> Subject: [ql-users] More about SMSQ/E v3.01 on SGC (and QXL) Hi again ! Well, I did more testing based on the feedback I received from you all (thanks ! That's what I like in the QL world: you always find an helpful hand and often many hands actually). Well, I did the following: - booting SMSQ/E v3.01 from a working v2.91 session: no ROM (Qubide, ROMdisq) intitialization, crash at the first keypress on the PC AT kbd (I guess it's because the interrupts are enabled at the hardware level on SH but that no driver is installed to process those interrupts). - unplugging the IDE cable from the CD-ROM drive (I got a hard disk as the master and the CD-ROM drive as the slave): no joy, SMSQ/E v3.01 crashing at load time (the screen stays in QL MODE 4 with flickering pixels at the bottom, denoting an activity of the OS (system variables updated at each polled interrupt). - unplugging the IDE cable from Qubide and booting from ROMdisq: SMSQ/E boots but here again, no ROM is initialized (i.e., after loaded from the ROMdisq under Minerva v1.92, SMSQ/E doesn't initialize the ROMdisq). - plugging back the IDE cable into Qubide and unplugging ROMdisq: Same as above: SMSQ/E boots but doesn't initialize the Qubide ROM once loaded. For info, I don't have the 'special GAL2' on Qubide (it is said in the ROMdisq manual that it won't work with a Qubide and a GoldCard if Qubide is not fitted with that special GAL, but I don't know if it extends to Super GoldCards as well). Also that system uses no backplane (Qubide is plugged into the Aurora and the SGC is plugged into the Qubide throug- connector). The power supply is a PC one, (200W, i.e. about 10 times what the system needs) and the +9V is derivated from the +12V via a LM317K regulator (with proper decoupling) so that Qubide and SGC are supplied via their own 7805 regulators (I found it the best solution to minimize the noise on power supply lines and the tension dropout from one card to another). The Qubide chimic capacitors were all replaced with tantalum ones (for a better filtering of high frequency noise) and the 74HC688 was replaced with a 74ALS688 (better rejection of noise too than it's CMOS equivalent). The resulting Qubide is rock solid and I never got a single hard disk corruption in 6 years of 24/7 functionning (this system was used for QLCF BBS), i.e. since the Aurora was installed (I did get problems with Qubide when the MB was still a QL one). I can't shorten the IDE cables (but it won't make a difference here, obviously). I conclude from the above that it must be a problem in the initialization code of SMSQ/E v2.98+ as everything works just fine with v2.91. I had a quick look to the sources, and if looks like SMSQ/E v3.01 checks for ROMs at addresses $C000 (ROM slot) and $4C0000 (inclusive) to $500000 (exclusive) with $4000 bytes (16Kb) increments. Following the instructions in the Aurora manual, my Qubide is configured to show its ROM from the base address of $4FC000, so it -should- be detected, like should be the ROMdisq (address $C000). BUT: I seem to remember that SGC and Aurora are using shadowing mechanisms, so my question is: are those addresses valid at boot time ? (Nasta, are you with us ? Three knocks for 'yes', two for 'no', please ! ;-) Another question is: what changed in the initialization routines (including cache handling ones) between v2.91 and v2.98 ? Using the v2.91 routines should allow to make SMSQ/E to work again (I don't beleive it's a hardware problem)... Tony (Tebby), are you around ?... :-) Another sad piece of news: while v2.98 works just fine on my QXLs, v3.01 don't and crashes after the boot process, apparently when a 'MODE' command is executed in my boot program... Marcel, was there a bug left in the display drivers when you sent the sources to Wolfgang ? Well, that's all for now... I was about to fiddle with the hard disk driver of SMSQ/E so to allow non-blocking ATAPI commands on my CD-ROM driver (there's a bug in SMSQ/E preventing to delay the ATAPI queue execution, for example when the CD-ROM spins up), but I'll wait until I find a way to have an homogoneous setting on at least both the Q60 and the SGC. QDOS/SMS forever ! Thierry.