GitLab

Previous changes were untagged, but incrementing the version number gives
us a tag we can associate with the official RISC OS 5.16 release.

Version 0.38. Tagged as 'Buffers-0_38'

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

Version 0.37. Not tagged

Bugfix to data copying routine: in the source unaligned, dest aligned case, an unnecessary word after the end of the source could be read, which is a problem if you're copying from the very end of a dynamic area, appslot etc.

Version 0.37. Tagged as 'Buffers-0_37'

Detail:
  A threshold value of -1 is meant to do a read, it was however also causing
an UpCall to occur because during 32bit conversion, the PSR was being
restored before checking for a flag that this was trashing.

  This bug affected 26bit builds as well and may have caused the odd
spurious bufferemptying/filling upcall.

Admin:
  Threshold UpCall no longer occurs when reading threshold. (Tested on RPC).
cd ~/RiscOS/Sources/HWSupport.

Version 0.36. Tagged as 'Buffers-0_36'

  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 0.35. Tagged as 'Buffers-0_35'

Detail:
  At a couple of points, the Buffer Manager would re-enable interrupts
    when it should not have done so.  These holes are now fixed.  Some
    other code has been optimised in light of extensions made to the
    CPU macros after this component was originally made 32-bit safe.
Admin:
  Built.

Version 0.34. Tagged as 'Buffers-0_34'

Detail:
  This has got to win some sort of award. The buffer manager service routine
  (reason code 7) and CnpV were _still_ getting the result wrong.

  Buffer_GetInfo (which uses a separate calculation) has been modified to
  ensure it returns the same result for free and used space when word-aligned
  buffers in use.

Version 0.33. Tagged as 'Buffers-0_33'

Detail:
  R14 really is spare, so that's used instead now.
Admin:
  Built (only affects 32-bit builds)

Version 0.32. Tagged as 'Buffers-0_32'

Admin:
  Phoebe & 32 machine builds tested.  Tested on Risc PC.
  Tested against old and new DeviceFS builds.

Version 0.31. Tagged as 'Buffers-0_31'

Detail:

  * See summary, above.  Basically the calculation of the free space
    was flawed.  Should now be fixed properly!

Admin:

  * Untested, but it's been verified by the BBVA (Bracey Bug-fix Verification
    Association)...

Version 0.30. Tagged as 'Buffers-0_30'

Detail:

  Added the missing "Buffer must be word aligned" error message to the
  German resources (although the sentence is in English).

  Non-internationalised version was missing the "Buffer must be word
  aligned" message; this has now been rectified.

  The buffer manager service routine "Return free space" (reason code
  number 7) was incorrectly calculating the free space.  Instead of
  using the size of the buffer and insertion point in the calculation
  it was using an uninitialised register and the reason code!  This
  was due to a change made in V0.27 to the stack environment which
  invalidated the usage of R0 and R1.  Registers R6 and R7 are now
  used in place of these.

Admin:

  Tested by softloading on an NC1 using the baseline Lazarus OS image.
  Fixes verified as being correct.


Version 0.29. Tagged as 'Buffers-0_29'

Version 0.28. Tagged as 'Buffers-0_28'

Version 0.27. Tagged as 'Buffers-0_27'

Changed to use srccommit.

Version 0.26. Tagged as 'Buffers-0_26'