GitLab

Admin:
  Required for recent libraries with improved amus.


Version 1.42. Tagged as 'PPP-1_42'

Version 1.41. Tagged as 'PPP-1_41'

Detail:
  Several flag variables were updated in Callback handlers
  as well as being accessed from foreground tasks, bu they were not
  declared as volatile.

  In practice, looking at the code it is unlikely that the variable would
  have been cached in a register when it was needed so this probably
  didn't cause any immediate problems.

Admin:
  Built. Unlikely to fix any recorded bugs.

  Note: Our compiler DOES support the volatile qualifier and any global
  variables that are updated in asynchronous routines MUST be declared as
  volatile. This includes any global variables used in Upcall, Callback,
  Event and interrupt handlers.

Version 1.40. Tagged as 'PPP-1_40'

Detail:
  Flags to indicate that callbacks were pending or that a callback was needed
  were not declared as volatile. This could potentially cause a problem
  as tey are set within interrupt handlers. In practice, looking at the code
  it is unlikely that the variable would have been cached in a register
  when it was needed so this probably didn't cause any immediate problems.

Admin:
  Tested in Customer F 8 soft load. This change doesn't fix any recorded bugs.

Version 1.39. Tagged as 'PPP-1_39'

PPP working with DebugLib.

Version 1.38. Not tagged

Version 1.38. Tagged as 'PPP-1_38'

Detail:
  serial_dcd was always returning 1.
  It now correctly returns the DCD state (1 when active, 0 when inactive).

Admin:
  Tested.

Version 1.37. Tagged as 'PPP-1_37'

Detail:
  * Now uses AsmUtils for interrupt control.
  * Uses new C system header files for definitions where possible.
  * Removed unused file hdr.CVars.
Admin:
  Blockdriver and DeviceFS variants built and briefly tested on a RiscPC
  by connecting to Demon and doing a bit of web browsing.

Version 1.36. Tagged as 'PPP-1_36'

Detail:
 	HdrSrc 0.46 has the macros defined in hdr.Cvars included. Hence have
altered s.ppp_io to use the version defined in Hdr:macros.
 	This change only affects the BlockDriver builds.

Admin:
	Built as both DualSerial and BlockDriver devices okay.

Version 1.35. Tagged as 'PPP-1_35'

Detail
    The FreeBSD changes mirrored in TCPIPLibs 5.24 require some extra header
    files to be included in places.
    Makefile 'clean' target now automatically strips dynamic dependencies.
Admin
    DeviceFS and Blockdriver variants compiled against TCPIPLibs 5.24.
    Requires TCPIPLibs 5.24 or later.
    Earlier versions of PPP will not build with TCPIPLibs 5.24 or later.

Version 1.34. Tagged as 'PPP-1_34'

Fixed a couple of bugs in SWI PPP_Status (reason code 1 - read physical line state), pointed out by Kevin Bracey.

Detail
    * Would return R2 undefined if the 'modem' flag was 0.  Now sets R2 to 0.

    * Dual Serial variant of serial_dcd routine used by this SWI was not
      checking whether the serial file handle was open before using it.
      It now does, and returns 0 if it's closed.
Summary
    Code compiled.

Version 1.33. Tagged as 'PPP-1_33'

Detail
    Once we've relinquished control of the serial port back to the Dialler,
    unset the 'using dialler' flag to ensure we don't act on any more Dialler
    Service calls.

    Altered debugging code to use DebugLib.
Admin
    DualSerial variant tested on a Risc PC and Customer M NC.
    Blockdriver variant untested.
    Removed svcprint.c and svcprint.h now we're using DebugLib.

Version 1.32. Tagged as 'PPP-1_32'

Extended list of options known to LCP packet dumper.
Added minimal CCP stub - mainly for printing packets. Will ConfRej all
protocols.

Version 1.31. Tagged as 'PPP-1_31'

1) Added a new reason code (1) to PPP_AlterSettings which not only sets the current idle time but also the default idle time.

This call is to be used when the ModemTimeout NVRAM location
is changed.
2) When either idle time is changed the zero the current idling
count. This prevents the situation where the idle time is lowered
but the count has already reached that point therefore the line
never drops.
  pppmodule.c changed and documentation in doc/BALChange updated.

Version 1.30. Tagged as 'PPP-1_30'

Details
    Service_PPPTimingOut was being sent directly from an interrupt handler, so
    if clients called non-reentrant SWIs the system could crash (e.g. NCMA
    playing a sound sample, leading to an abort in DeviceFS on Customer M NCs).
    Now sent from a callback handler instead.  Also fixed problem where
    Service_PPPTimingOut could have been sent more than once (and the
    connection termination code called more than once) if the first connection
    termination took >= 1 second.
Admin
    Source compiled.  Tested on an Customer M NC.

Version 1.29. Tagged as 'PPP-1_29'

Now dynamicaaly creates and deletes the rm and aof directories. This allows it to build straight out of CVS.

Version 1.28. Tagged as 'PPP-1_28'

Detail:
    The fix in the previous to reset serial_in and serial_out correctly on
    DualSerial build was not conditionally wrapped by "#ifndef BLOCKDRIVERS",
    so the code was being included blockdriver variant and since the variables
    don't exist the link would fail.
Admin:
    Sources compiled.

Version 1.27. Tagged as 'PPP-1_27'

CMHG, but somehow seems to have been in CVS all along without getting
noticed.

Version number not incremented.

Version 1.26. Not tagged

Function serial_finalise resets serial_in and serial_out to 0 if we're NOT
using the dialler, but if the dialler is being used they weren't being reset.
This meant that typing pppinfo on the command line only reached as far as
serial device.  The next bit of code depended on serial_in, which when it
hadn't been reset, wrongly resulted in OS_Args being called with a duff
handle for serial_in.  OS_Args lives in fileswitch, which is where the abort
was showing up.  Haven't been able to repeat the abort since this fix.  Fix
also verified by Ben.

Version 1.26. Tagged as 'PPP-1_26'

    Removed some debugging information that had slipped through the net.

Version 1.24. Tagged as 'PPP-1_24'

Changed flags which prevent multiple callbacks to serial port data reading routine in DeviceFS build of module due to fears that data input may call

SWIs which allow more callbacks to go off, hence causing a reentrancy problem.

Version 1.23. Tagged as 'PPP-1_23'

Screwed up previous check-in: checked in an experimental version of serial.c by mistake instead of the 'proper' version.

Version 1.22. Not tagged

Version 1.22. Tagged as 'PPP-1_22'

Fixed bug where callback to read data would not be removed, causing PPP to block heavily in certain situations.

Version 1.21. Tagged as 'PPP-1_21'

Added MemCheck build capability.  Found and fixed a significant memory leak as a result of this, which would lose you whatever your MRU size was plus about another 130 bytes (for headers, FCS etc.) per pppconnect.

Version 1.20. Tagged as 'PPP-1_20'

Version 1.19. Not tagged

Version 1.19. Not tagged

Got rid of a few more compilation warnings.

Added diallerb option to use DiallerB for serial stream handle
arbitration instead of dialler.

Removed ability for PPP to use Dialler/DiallerB for timeouts and line
drop detection - nothing used these and they have no advantages over
PPP's internal routines for doing the same.

Fixed bug in *command handler (overwriting command tail).  Got rid of a couple of warnings.  Added a doc directory containing some documentation and a history file.  See history file for more details.