GitLab

When presented with a module, squeeze will squeeze it and tack on a header which does OS_Module 11 on it when unsqueezed.
However, the header size (bytes) was being added to a pointer (of type int) leading to the header being loaded 4x too high, overwriting the heap - size now adjusted.

Also remove the assumption that clock() counts in cs, so that the times come out right on a cross compiled Unix system.

Version 5.14. Tagged as 'squeeze-5_14'

Detail:
  * Reinstated Desc to INSTAPP_FILES
  * Changed path to Env directory, assuming source trees are no longer merged

Retagged as 'squeeze-5_13'

Version 5.13. Tagged as 'squeeze-5_13'

CApp does this for us.

Version 5.12. Tagged as 'squeeze-5_12'

Detail:
  Gory details are in a long source file comment, but basically this does
  what's necessary to coax RISCOS Ltd's version of the UnsqueezeAIF module
  into letting compressed binaries run again, without sacrificing the cache
  coherency bugfixes in later versions of squeeze.
Admin:
  Reportedly now generates executables that work on an A9Home.

Version 5.11. Tagged as 'squeeze-5_11'

Retagged as 'squeeze-5_10'

Detail:
  The code was continuing to reference a couple of data blocks after they
  had been freed, leading to corrupted output files that failed with
  undefined instruction errors. Potentially could have affected RISC OS,
  but this was first noticed on Linux where the failure seems to happen
  every time.
Admin:
  Verified that it now produces correct executables.

Version 5.10. Tagged as 'squeeze-5_10'

Detail:
  * Added shell-script make wrapper
  * Uses shared makefile fragments and one master makefile for all variants
  * #includes use CLX path explicitly
  * SWI numbers defined explicitly in s.unsqueeze to avoid dealing with the
    issue of RISC OS paths in GET statements
Admin:
   No functional change, so version not incremented

Tagged as 'squeeze-5_09-2'

Version 5.09. Not tagged

Detail:
  Gave Components file more power to install the executable and frontend in
  different places. Iyonix-themed application directory sprites are now
  available in addition to the old Morris4 ones, and are selected based upon
  the UserIF build variable.
Admin:
  Checked frontend installs correctly.

Retagged as 'squeeze-5_09'

Came to light in Rhenium; have been unable to reproduce in Iyonix.

It may be worth updating UnsqueezeAIF to identify the incorrect unsqueeze
code, and re-adding UnsqueezeAIF to the Iyonix ROM (as well as issuing as
a soft-load update for StrongARM Risc PCs).

Version 5.09. Tagged as 'squeeze-5_09'

xpand would correctly decompress old non-AIF images, and the broken non-AIF
images generated by recent versions of squeeze, but needed to be changed to
also decompress the new non-AIF images.
Memory limit check now always occurs if running in the application slot,
regardless of whether we were loaded at &8000 - the check was not performed
reliably for non-AIF images either.

Version 5.08. Tagged as 'squeeze-5_08'

Detail:
  If squeeze completes without the value from clock() changing, the
    rate calculator ends up dividing by zero!  Now trapped.
Admin:
  Not built, not tested - will fix the problem.
  Reported by David Ruck on c.s.a.programmer.


Version 5.07. Tagged as 'squeeze-5_07'

Detail:
  The insertion of the SWI XOS_SynchroniseCodeAreas in the squeeze code
    defeated xpand's attempts to decipher the code.  xpand is now willing
    to look back up to 8 instructions for the SUB r8, r8, #&7C instruction.
Admin:
  Tested - xpand can now expand things squeezed by squeeze again.

Version 5.06. Tagged as 'squeeze-5_06'

Version 5.05. Tagged as 'squeeze-5_05'

  Updated !Run file to ensure the latest modules.
Admin:
  Requires BuildSys 3.01 or later.
  Co-dependent with Library 0.67 or later

Version 5.04. Tagged as 'squeeze-5_04'

Detail:
  The code inserted by squeeze into an application to perform
    unsqueezing is now StrongARM compatible and no longer relies
    on the presence of the UnsqueezeAIF module.  Since it has
    better information on what it's been updating, it can actually
    make a superior call to OS_SynchroniseCodeAreas marking the
    exact area of memory that it has updated.
Admin:
  Tested on Risc PC with UnsqueezeAIF unplugged.

Version 5.03. Tagged as 'squeeze-5_03'

Detail:
  Added install: rule to Makefile to build FrontEnd UI.
Admin:
  Tested on RiscPC

Version 5.02. Not tagged

  Fixed filetype failure (RISC OS build)
Detail:
  If the application memory limit for a squeezed application was large
    enough to hold the squeezed code but not large enough to hold the
    unsqueezed version without overwriting the calling application
    (eg.  the C compiler being too large to fit under amu), then
    everything would die nastily.  This fix to the unsqueeze code
    verifies that the calling program will not be overwritten and
    aborts the unsqueeze operation before disaster strikes.
  RISC OS builds remember to retype the file after saving.
Admin:
  Tested by forcing amu to launch a newly-squeezed cc image in the
    appropriate memory conditions, and cc generates an error before
    unsqueezing and killing amu.
  Verified that UnSqueezeAIF still recognises and patches the unsqueeze
    code for StrongARM compatibility.

Version 5.02. Tagged as 'squeeze-5_02'

You need CLX and aoftoc compiled for your architecture for it to
work. To do this converted the calls to _kernel functions into calls
to the wf_ functions in CLX.
xpand hasnt been ported since it currently works by running the
unsqueeze code that squeeze inserted into the binary and this
doesn't work too well on non-ARM architectures.

Version 5.01. Tagged as 'squeeze-5_01'

  Also contains sources to xpand tool.
Detail:
  Updated to take version numbers and dates from VersionNum
    but otherwise unchanged.
Admin:
  Changed to use srccommit.
  Initial VersionNum file faked to represent current version.

Tagged as squeeze-5_00