GitLab

Detail:
* Makefile: get directory separator and header extension from shared makefiles
* Use ResourceFile macro directly for standalone case
* Enable GitLab CI

Version 0.74. Tagged as 'DeviceFS-0_74'

Detail:
  The Unlink macro also released the memory back to the RMA, but in 2 cases passed the old pointer around in a service call. Change Unlink to only remove from the linked list, and add a Free macro to do the release after the service call.
Admin:
  Spotted by Colin Granville.

Version 0.73. Tagged as 'DeviceFS-0_73'

PRM 2-612 says the device driver can modify R3-R6 during buffer creation in order to change the buffer creation parameters used with BufferManager.
However, the subsequent logic in DeviceFS meant that if the buffer already existed the flags (R3) never got used.
In practice this only affects the magic buffer numbers 0-9 retained for BBC Micro compatibility (Keyboard/Serial in/Serial out/Printer/Sound 0-3/Speech/Mouse), and of those only Serial & Parallel are in DeviceFS, and of those Parallel had a workaround, so really only the SerialDeviceDriver gains a fix here.

Tested in an IOMD ROM.

Version 0.72. Tagged as 'DeviceFS-0_72'

Detail:
  s/DeviceFS - Escape some dollars contained in strings to avoid warnings from objasm
Admin:
  Resulting binary unchanged


Version 0.71. Retagged as 'DeviceFS-0_71'

An early PullEnv corrupted R0 so the tests of ff_FileError and ff_TimedOut were on the wrong flags.

Version 0.71. Tagged as 'DeviceFS-0_71'

The DeviceFS API is defined in terms of actual useable buffer sizes, but BufferManager only makes use of n-1 of those requested of it (or n-4 for word bnuffers) per PRM4-88.
This change adjusts the request so the client of DeviceFS gets the useable number of bytes desired.
Changes from Colin, ref http://www.riscosopen.org/forum/forums/3/topics/1895

Version 0.70. Tagged as 'DeviceFS-0_70'

Same aim as previous commit, but keeping the default as 'block'.
Retagged as DeviceFS-0_69.

FSystem.s:
* the open[in|out] operations cleared any flags that HandleSpecial collected
* reversed the sense of [no]block
Special.s:
* rewind the input pointer too, otherwise parameters must be in the same ascending order as the validation string

Ref http://www.riscosopen.org/forum/forums/3/topics/1888

Version 0.69. Tagged as 'DeviceFS-0_69'

Detail:
  deregisterdev and CallDevice now take care not to load from an unaligned address if the NoUnaligned option is turned on.
Admin:
  Tested on rev C2 beagleboard


Version 0.68. Tagged as 'DeviceFS-0_68'

Detail:
  Uses suffixed file extensions for compatiblity with both objasm and asasm.
Admin:
  Supplied by Peter Naulls, tested at ROOL

Version 0.67. Not tagged

Detail:
  s.FSystem
 line 1515 moved to approx 1460.. where file_InternalHandle is still valid
Admin:
 Tested in ROM at Castle


Version 0.67. Tagged as 'DeviceFS-0_67'

Detail:
  Changed s.FSystem
Admin:
  Updated from Kev's machine by Mike H.

Version 0.66. Tagged as 'DeviceFS-0_66'

Now ensures that the device driver WakeUpForRX entry is always called with the exact byte count in R3, even when DeviceFS was entered via the

FSEntry_GetBytes reason code. (Previously, R3 held too large a value in
this case.)

Version 0.65. Tagged as 'DeviceFS-0_65'

A signed string pointer comparison was stopping USBdriver from finding valid endpoints when ROMmed,as the special string it registers with was at a -ve address.

Changed to behave per the PRM (only 0 is important).
That was a git to find!

Version 0.64. Tagged as 'DeviceFS-0_64'

The docs say if you DeviceFS_Register with your file handle limits set to minus one then there's no limit,however this was never being checked for.

Now it does.

Version 0.63. Tagged as 'DeviceFS-0_63'

Where there is more than one stream open on a device the process
of doing a directory listing corrupts the attribute flags due to a
MOVS r3, r4 where the routine later needed to use r3.

Version 0.62. Tagged as 'DeviceFS-0_62'

Detail:
  Debug switch missing
Admin:
  It does now build, but no other testing carried out.

Version 0.61. Tagged as 'DeviceFS-0_61'

Detail:
Fixed by Dan Ellis
Admin:
He tested it, and it worked

Version 0.60. Tagged as 'DeviceFS-0_60'

  - Fixed sleep and timeout ioctls
  - Fixed initialisation from special field

Detail:
  OS_RemoveTicker was being called with the wrong handle so not being
removed.  This was causing timeout behaviour to work incorrectly.

  Sleep and Timeout ioctls weren't reading their reason fields from the ioctl
block.  They now do, and work!

  Defaults weren't being set correctly from the special field.

Admin:
  Tested on a RiscPC, and STB.

  Timeouts cause an error and so because of the fileswitch interface will not
actually return bytes read during that request.  To perform a managed timeout
in which bytes can be counted during timeout, non-blocking calls must be
made.  However, bytes have still been copied to the destination buffer
(although the caller won't know how many).

  There are still 'issues' with special fields.  However, they appear to work
correctly on the Dual Serial interface.



Version 0.59. Tagged as 'DeviceFS-0_59'

Detail:
  DeviceFS has been enhanced with the ability to cooperate with Taskwindows
so as to allow multitasking to continue.  Any devicefs stream, if opened
with the special field entry 'sleep' will issue OS_UpCall 6 when it would
normally block.  A flag 'timeout' has also been added which will cause the
stream to issue a timeout error if it has slept for too long.

  These actions can be set using the generic IOCTLs 4 and 5, the former just
being a logical switch, the latter being the value in centiseconds for which
to sleep.

  There is also now a generic special field to specify non-blocking which may
be used as an alternative to the IOCTL.

Admin:
  Tested on a RiscPC.


Version 0.58. Tagged as 'DeviceFS-0_58'

  Updated to build using objasm instead of aasm.
  Sources changed to be objasm-compatible.
Admin:
  Requires Library 0.71 or later.
  Requires BuildSys 3.06 or later.
  Requires Env 0.65 or later.

Version 0.57. Tagged as 'DeviceFS-0_57'

Detail:

  * Exports C header file required by SoftIR module.

Admin:

  * Built on RPC.  SoftIR module compiled using the newly exported header.

Version 0.56. Tagged as 'DeviceFS-0_56'

Detail:
  lr was being used as a temporary register to store the buffer handle in
whilst checking for ownership.  If DeviceFS didn't own the buffer the call
was being made on, it then used this value in a MOV pc, lr thus going who
knows where...

Admin:
  Checked and still works.  Appears to cure previous problem where issuing
Upcall 8 could cause a branch through zero error.

Version 0.55. Tagged as 'DeviceFS-0_55'

Version 0.54. Tagged as 'DeviceFS-0_54'

Version 0.53. Tagged as 'DeviceFS-0_53'

Bug fix: DeviceFS was inadvertently enabling interrupts when generating the Upcall to indicate that new data has arrived in an input buffer.

Detail:
  DeviceFS was using OS_UpCall to generate a UpCall_DeviceRxDataPresent
  upcall for for the wake up event (when data was inserted into a dormant
  receive buffer).

  However, OS_UpCall enables interrupts, contrary to the RISC OS 3 PRM.
  Since it happened within the hardware modem interrupt handler, this is
  quite dangerous and can lead to the modem interrupt handler being
  re-entered, thereby corrupting bytes or putting bytes into the buffer in
  the wrong order.

Admin:
  SCTransport v0.57 or later is needed if using this version of DeviceFS
  (since earlier versions relied on this incorrect behaviour in DeviceVS).

  Unfortunately, Pace fault 2098 still exists after making this change
  (Chip to Server transactions are very slow in Customer F 8 using a HW modem).

Version 0.52. Tagged as 'DeviceFS-0_52'

Detail:
  The UpCall handler was using MOVEQS pc, lr to pass on the vector in
    some cases.
  Fixed a non-internationalised build bug - missing condition in macro.
Admin:
  Built.

Version 0.51. Tagged as 'DeviceFS-0_51'

  Added stripdepnd to MkClean
Detail:
  Replaceed -throwback with ${THROWBACK} so that throwback isn't generated
  in ROM builds, but still works for local test builds.
Admin:
  Tested on Ursula build.

Version 0.50. Tagged as 'DeviceFS-0_50'

Version 0.49. Tagged as 'DeviceFS-0_49'

Some stray border flash debugging removed. FSEntry_Args code for extent was broken in the non-FastBufferMan case (not that this will ever have been used).

Version 0.48. Tagged as 'DeviceFS-0_48'

Detail:
  Also fixed faulty debug code which tried to report the escape state,
    but would have always failed to do.
Admin:
  Tested Phoebe and 32 machine builds.  Tested on Risc PC.

Version 0.47. Tagged as 'DeviceFS-0_47'

Version 0.46. Tagged as 'DeviceFS-0_46'

Added purging of output buffer if checkfileTXOK returns an error to get round the problems of paper out on lexmark printers.

Version 0.45. Tagged as 'DeviceFS-0_45'

German resources imported. Capitalisation of "IOCtl" fixed.

Version 0.44. Tagged as 'DeviceFS-0_44'

Changed to pass device driver stream handle to DeviceDriver_IOCtl, rather than the DeviceFS stream handle. WARNING! Anyone that was relying on the broken behaviour will have to change their code.

Version 0.43. Tagged as 'DeviceFS-0_43'

Version 0.42. Tagged as 'DeviceFS-0_42'

Changed to use srccommit.

Version 0.41. Tagged as 'DeviceFS-0_41'