-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathbugs.txt
98 lines (70 loc) · 3.13 KB
/
bugs.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
$Id: bugs.txt,v 1.6 2019/10/18 18:53:50 albert Exp $
*1.Make gsetalpha is different in 32 and 64 bit mode.
Almost nothing works in 64 bits.
Solution: log of 2019 oct 14 & oct 16
*2. J, gives a completion containing n| which is wrong.
Observation: n| sets the TALLY-GO to 0.
The reason show-completion did not save/restore TALLY-GO.
-----------------------------------------------------------
OLD RANTS.
Unsolved:
In the documentation there is a need to warn for redefinitions.
The -dc- thingies should be generated with a automatic tool.
MAKE-CUL should print symbolic names for ranges where possible
There is some worry because QSORT can't handle empty ranges
(or so it seems).
A lone EQU leads to a not unique warning in the second pass.
If a label is defined using EQU the not-unique-message
should be supressed.
Byte offsets should be signed. Maybe all offsets should be signed.
Introduce AS-@+ .
ORG is a mess. It clears the code buffer, so it should only be issued
at the start of a segment. -ORG- changes the association with the
start of the segment. This is a bad idea, it just means that we are
entering a different segment.
We need a don't care segment that is not analysed, because it contains
don't care items.
WHERE-LABEL and CLEAN-LABEL require sortedness. This should be a permanent
and maintained property such that the use cannot be accidentally wrong.
The name SEGMENT is all wrong. It should be called SECTION !!
------------------------ WISH-LIST ---------------------------------
There could be a priority to cul-items. If there is an overruling
priority the lower priority would give way.
Example: disassembling at priority 1 hits a data area with priority
0. The disassembling stop there.
Disassembling at priority 1 would stop at an unknown instruction,
at priority 0 it would not.
------------------------ REPORTED ---------------------------------
----------------------- AFTER 2018 ----------------------
Confirmed per 2018 sep 4
The 8080 assembler allows things as
MOV M'| M|
PUSH SP
LDAX HL
because no BAD bit at all are used in the thingies.
Confirmed per 2018 sep 4
4.53
Completion now works generally ok with one exception:
[MEM expects tally bits surviving the SIB instruction,
but after `` LEA, AX'| ZO| [MEM '' SHOW-COMPLETION
doesn't work. ( error 28)
This looks most like a bug in the 386 assembler than the
generic system.
Not so easy involves sib-byte.
Rejected per 2018 sep 11
There is a problem around the disassembly of variables like
base and h0 in rf751.asm
rf751 gives a phase error on tib. This can be ignored.
It is not understood either. It has nothing to do with the
existance of a word TIB in the kernel of Forth.
.
The greatest problem is that in fact the format of the
data in the PIFU's is not documented.
As per 2019 the J, no longer disassembles correctly.
Bug in 64 bits ciasdis only:
As per 2019 oct 15 in the test example, the disassembly doesn't
try to find a nearby label. So START 1 + is a diff.
As per 2019 oct 15 in the test example, on the 64 bit system
the rf751 example fails, because of V_base not defined.
The problem is identified as to be with the crawling of
rf751, so restricted to rfcrawl.cul