[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]

michael/wjr merge

The frozen version of the merge of my stuff with wjr's is in

The human-generated files are frozen.  I may delete the machine-generated
files to save space.

My starting point was alpha-5.  Wjr's starting point was somewhere
between alpha-5 and alpha-6, and the reference point has been lost.
Some files from the frozen version of alpha-6 have been included.

wjr, hibbert, and I are now using this as the reference point from
which the three of us diverge.

(I note that three people working on the same chunk of code, diverging
from a point already diverged from the main line (where at least one
more is working on related chunks) should make the next merge considerably
more interesting.)

The change and regression-testing status are as follows:


  The source files in this directory should be only those that are
  automatically translated from Smalltalk, without human-keyed changes.
  Wjr does not recall making any changes since he grabbed the
  between-alpha-5-and-alpha-6 code.

  The makefile provided did not, and could not easily, run regression
  tests.  I added a "tests" target and ran them.  They all failed.
  I have begun pursuing this with the Smalltalk crew, and will
  continue over the next few days.

  I also plan to write a letter explaining the pattern of use I
  intended for the regression test support.


  Minor but significant changes were made.

  The test passes to the same extent it did before my hacking began.  I.e.:

   - Though in a file whose name implies it is a regression test, the
     test is actually a test-prod test, which tests only one aspect of
     the module.  (This is not an entirely unreasonable confusion, since
     we don't yet have a name-extension defined for test-prod tests, which
     I though we wouldn't be using.)

   - The same reference file is used to test both the integrated and
     server/client versions.  The integrated version passes, the
     server/client version shows one difference: a change in the
     report of a memory-management item in a message from the bomb
     package.  I have not examined whether the difference is a bug
     or a proper manefestation of the server/client mode.


  Extensive and significant changes were made.  Merging this section
  with the alpha-sequence will be easier if I am present.

  All regression tests pass except the one for init.  Init and its test
  are from alpha-6.  The init regression test is known to be incomplete
  and the failure is the properly expected behaivor.


  Extensive and significant changes were made.  Merging this section
  with the alpha-sequence will be easier if I am present.
  There is currently no regression test for stubble.  Its correctness
  might be inferred from correct operation of complete tests of the
  garbage collector, communication (whose testbed is calc), and snarf
  tests.  (Note that calc does not yet have a regression test, com is
  in flux, and the cook-count on this soup has just climbed to three,
  while snarf is still under construction.)