Thiemo Seufer wrote:
> Brian Mastenbrook wrote:
> > Thiemo Seufer wrote:
> > >I had a closer look and found a lot more bugs than expected. The
> > >resulting diff against CVS HEAD is >50k and improves stability of
> > >the mips port significantly. I'll split it in pieces and add a short
> > >explanation to each.
> > Wow. Nice work. A few questions:
> > Does the resulting SBCL build itself and pass its tests?
> I did several build-bootstrap cycles with 0.8.21.16, the patch against
> CVS is nearly identical to what I used there. The non-passing tests
> remained consistent, the last part of the patch series excludes the
> failing bits for now. Most of them are also problematic on other
Tested with a full cycle of 0.8.21.50 + my patches now, no changes.
> > Have you run the ANSI tests on it?
> No, I concentrated on stopping it to crash first. :-)
I ran the daily-test framework in CVS for it:
69 out of 20332 total tests failed: DOLIST.17, CALL-NEXT-METHOD.ERROR.1,
CALL-NEXT-METHOD.ERROR.2, ATANH.4, ATANH.5A, /.ERROR.5, /.ERROR.6, /.ERROR.7,
/.ERROR.8, SUBTYPEP-FUNCTION.1, SUBTYPEP-FUNCTION.2, SUBTYPEP-FUNCTION.3,
SUBTYPEP-FUNCTION.4, SUBTYPEP-COMPLEX.8, PATHNAME-HOST.9,
LOGICAL-PATHNAME.ERROR.2, LOGICAL-PATHNAME.ERROR.10, READ-SEQUENCE.ERROR.7,
WRITE-SEQUENCE.ERROR.3, FILE-LENGTH.ERROR.8, MAKE-BROADCAST-STREAM.5,
MAKE-BROADCAST-STREAM.7, MAKE-BROADCAST-STREAM.8, WITH-OUTPUT-TO-STRING.16,
PRINT.BACKQUOTE.RANDOM.6, PRINT.BACKQUOTE.RANDOM.7, PRINT.BACKQUOTE.RANDOM.8,
PRINT.BACKQUOTE.RANDOM.14, PPRINT-FILL.14, PPRINT-FILL.15, PPRINT-LINEAR.14,
PPRINT-TABULAR.13, PPRINT-LOGICAL-BLOCK.17, PPRINT-POP.7, PPRINT-POP.8,
PPRINT-POP.9, PRINT-LEVEL.2, PRINT-LEVEL.6, FORMAT.LOGICAL-BLOCK.CIRCLE.1,
READ-SUPPRESS.14, READ-SUPPRESS.15, SYNTAX.DOT-ERROR.7, COMPILE-FILE.2,
COMPILE-FILE.13, COMPILE-FILE.14, COMPILE-FILE.16, LOAD-PATHNAME.1,
LOAD-TRUENAME.1, APROPOS.ERROR.2, APROPOS-LIST.ERROR.2, MISC.63, MISC.64,
MISC.65, MISC.89, MISC.89A, MISC.89B, MISC.293A, MISC.363, MISC.414,
MISC.439, MISC.460, MISC.534, MISC.555, MISC.580, MISC.596.
2119.976 seconds of real time
1935.53 seconds of user run time
184.12 seconds of system run time
2127 page faults and
13,623,637,392 bytes consed.
Not that bad when compared with x86-64. In ansi-tests-compiled.log those
DOLIST.17, CALL-NEXT-METHOD.ERROR.1, CALL-NEXT-METHOD.ERROR.2, ATANH.4,
ATANH.5A, /.ERROR.5, /.ERROR.6, /.ERROR.7, /.ERROR.8, COPY-SEQ.19,
Lateron in that testsuite it fell in ldb with some gc weirdness:
STRUCT-TEST-65/3 STRUCT-TEST-65/4 STRUCT-TEST-65/5set_auto_gc_trigger: tried to set gc trigger too high! (0x04673e28)
fatal error encountered in SBCL pid 3201:
<Frame 0x000000b8, CODE: 0x0188B58F, COLLECT-GARBAGE, LRA: 0x0188b58f, PC: 0xffffffe8>
Bogus callee value (0x000000b8).
Really quit? [y]
The tests after that weren't run.