Meeting Pearls 3/Meeting_Pearls_III.iso/Pearls/bench/SeekSpeed/History.doc

Text File | 1993-05-22 | 7KB | 206 lines

37.12    22 May 93

  Change to include Emulation code result for 68040 FPU was
  overwriting CPU code string. Strings incresed to 50 chars. Now
  finally correctly reports all CPU's and FPU's.


37.11    4 May 93

  Added exit(0) at termination. Was returning random result code. I
  think this was reported, but I forgot it. Thanks to Eddy Carroll.

  It appears some CD's can have seek times in excess of 1.2 seconds.
  Changed the result string gadget format to conditionally test and
  print as "%6.1f" if >999.0mS, else "%6.2f" . Thanks Jez and Dave.

  Bug fixed with FPU detection. FPU was being wrongly reported due to
  wrong assumptions about 040 flag. Now correctly reports all FPU's
  and advises if 040 FPU is running on it's own or with the 68040
  library Emulation code loaded. Thanks sclifton@cix and
  kholland@cix.



37.10    25 April 93   (First release 2 May 93)

  Added name of Device Driver to Proforma

  Corrected spelling of picky Beta testers name in docs :-)

  Removed debugging printf()'s and window, as IDE bug now fixed

  Added error trapping and abort. If errors occur during a test run,
  the error will be reported and the remainder of the tests aborted.

  Tidied up some resource allocation. After a real thrashing, she
  gave back every single byte !

  IO buffer now long aligned, thanks Steve.

  No longer detaches from CLI. Cback.o was causing problems, still
  not recognising Ctrl-C to abort test though

  OK, I've given up trying to abort a test with Ctrl-C or whatever.
  Everything I've tried has failed. So we are linking with cback.o
  again, and it detaches from the CLI again.



37.9    17 April 93

  Added some debug code to try find the IDE problem

  Replaced SendIO() / WaitIO() with DoIO

  Now correctly passes Startup->fssm_Flags to OpenDevice(), was always
  opening the device driver with Flags = -1

  Now correctly gets the BytesPerSector from the Info() command,
  rather than assuming 512

  Now correctly gets the total sectors from Info() command, rather
  than calculating from the Environ array. This correctly reflects
  the reserved blocks.

  Device Requester now opens with NULL as default device. Default was
  previous garbage, and gave rise to "Device <$%*&*&> not found"
  requester. YARP (yet another reqtools problem). If the user hits
  return without entering a device, the requester is cancelled. If
  the user clicks OK without entering a device, the requester passes
  a NULL string back as a valid device !

  Now uses BytesPerSector for the tests. Should theoretically work on
  devices with other than 512bps


37.8    16 April 93

  Added requester after clicking save to give the choice between Save
  or Append

  All results now column alligned

  Changed process name to SeekSpeed_Process. Original, huh ?

  Changed a few requesters that won't get seen unless there's an
  error

  Finally got the window locking and BusyPointer working on all
  requesters

  Appears to be major BUG, in that she fails on IDE drives with
  invalid device and gets confused. This could be volume name is
  being returned, but with DLT_DEVICE set. Need to select Devices
  only

  Removed string gadget for entering Device Name and replaced with
  Text gadget to display it and Device requester to get it. Seems to
  work OK. Should be able to re-enable START button now, as device is
  still vailid at end of run.

  After testing a valid device, the START button is left enabled so
  you can re-test over and over again if you wish

  Return from AddPart was being wrongly assigned to a global which is
  one of the loop counters in the test routine. Global/Local mixup.
  May have been causing some strange results

  Tidied up a lot of requesters. Now correctly reports if results
  have been saved or appended, or if append or save has been
  cancelled


37.7    12 April 93

  Added DateStamp to Proforma

  SeekSpeed now detaches itself when 'run' from the CLI

  *BUG* (to be fixed) If you do more than one test, then memory is
  lost. The cleanup code is a big change, as it is only effectively
  freeing some of the structures and messageports from one test.
  Think I'll leave this one until later :-)

  Some (most) of the above fixed. Each time a Device is entered, The
  MessagePort and IORequest are freed, NULLed, and then new ones
  assigned to the new device. This results in Start being ghosted
  until a Device is entered, so you can't test again until
  re-entering the Device name. Still have to fix a possible missing
  CloseDevice() if more than one test is conducted

  Device is now 'Busied' during the test. I personally don't like
  this, as it takes longer for the system to implement than it does
  to test a ram drive, and if the HD is busied, all windows are
  closed, the backdrop cleared, and if, during the test, you click on
  the Icon for the assigned volume name, it asks you to insert the
  Device: . Comments please Beta testers ?

  Later... I've compromised and only Busy the device if it is DFx: .
  Floppies take a long while, and can safely be busied. The effects
  of Busying the HD were horrible

  Now fixed the unclosed Device. All memory is returned, and all
  messageports, IORequests and Devices safely closed down, on exit,
  or when choosing a new device. Please check this well Beta guys. No
  changes to the manual since V37.6


37.6   11/12 April 93

  Added file requester for saving results (at last...)

  Added ProForma information to Results

  Added CPU and FPU auto-detection to Proforma

  Added AddBuffers detection to Proforma

  Removed gadget ghosting, and added BusyPointer during SAVE and
  PRINT

  Added 'File saved' confirmation requester.

  Removed disasterous bug that was waiting to happen. Managed to get
  two nested rtLockWindow/rtUnlockWindow. If the middle one had
  happened (unable to open file/printer) then unallocated memory
  would have been freed.


37.5  10 April 93

  Guru when exiting after choosing Ram: . Although the device is
  found in the dev list and locked, it is correctly detected as not a
  valid device for this test. However disk info is assigned. When
  exiting the DoIO had invalid data and crashed. Only way round this
  was to do a string comparison for ram: and never look at the unit.

  Added the letters T, F, B for total, free and bytes/sector into the
  device gadgets.

  Now uses ReqTools Library for requesters, and for saying Ram device
  cannot be tested. Also added BusyPointers where needed

  Added Menus with Quit and About items. Made them NewLookMenus


37.4  9 April 93

  Fixed the memory leakage. Few FreeMem()'s missing !

  Re-worked the algorithm for number of timing events used. Now default
  to 100, but each test call will dynamically calculate the maximum
  allowed number of tests and set n accordingly, such that the
  device is never over-seeked.

  Print and Save functions implemented.


37.3  8 April 93

  First test vesion. Functioning, but some memory leakage. If device
  has low sector count, it can seek past the end.