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
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
|
Name : SocketWatch
Purpose: Asynchronous socket operations without null polls
Version: 0.07 (11-01-2019)
Status : GNU GPL
Authors: Dickon Hood, Frank de Bruijn
This isn't an official release.
In November 2003 I noticed SocketWatch 0.04y had a couple of bugs. After I had
corrected them and was almost ready to release 0.05, I discovered the
'official' place for SocketWatch was now on the Nettle site
(http://nettle.sourceforge.net). So I contacted the Nettle maintainers, told
them about the bugs and sent them all the work I had done on repairing it,
more or less expecting to hear something about an official release 0.05 within
a couple of months...
Still nothing three months later, so I sent one more e-mail (which remained
unanswered) and put 0.05 on my site.
Installing
==========
The zip archive contains an application called !ModInst. Run it to install the
module on your computer. If you want to install the module manually, you can
find it in !ModInst.Resources.!System.310.Modules.Network.
Source
======
The source code is provided as an ObjAsm file in the Source directory in the
archive. If you want to assemble the module again, run the Build file. The
module will be recreated and moved to the !ModInst application for
installation.
Version history
===============
11-01-2019 0.07
----------------
Catered for errors in SocketWatch_Register better.
Added notes about use with AcornSSL.
12-06-2012
----------
Removed dependency on private library (providing constants) so the source can
be used stand alone. Added Build script (previous builds were done from
StrongED, using a multipurpose ObjAsmRun script). Nothing changed otherwise.
MD5 checksum still f2bb3f8d911d0b20a43d6ee135c3ae97.
23-06-2006 0.06
----------------
Fixed small bug in the AtomicReset SWI (flags not restored in 26 bit mode, so
IRQ bit not cleared). Not sure if this ever bothered anyone (think not as it
was already present in 0.04y).
Removed disabling interrupts in EventHandler as event handlers are entered
with interrupts disabled.
Also some more source clean ups and more comments added.
MD5 checksum: f2bb3f8d911d0b20a43d6ee135c3ae97.
14-08-2005
----------------
Moved the source to ObjAsm. No functional changes to the code. MD5 checksum
still 90f3323981a0e287f488e9a59b2e5373. No new release.
12-12-2003 0.05
----------------
First one by me. Came about when I was investigating a memory leak on my RPC.
MD5 checksum of the module: 90f3323981a0e287f488e9a59b2e5373.
The detailed list of changes below was for the benefit of the Nettle
maintainers and originally sent to Ian in an e-mail on 11-12-2003.
Use the single file source of 0.04y as a reference.
- Most of the variable definitions from Flags% to r12End% aren't required. You
just need SockList% (with a FNdb(-4) instead of a FNdb(4)) and r12End%. The
rest was removed.
- No flag definitions are needed. Removed.
- Socket structure definitions and ErrorChunk%: no changes.
- From the module header to EventHandler_GotOne32%: merely some optimisations.
- Error texts: were placed near the routines that use them. This was mainly
cosmetic, but it had the added advantage of allowing me to replace all the
FNADRLs with simple ADRs.
- GetBlock%: removed.
- ReleaseBlock%: removed.
- IsHexNum_lp1%: not required. Removed.
- moveBytes%: not required. Removed.
- From title string to SWI-handler: no real changes.
- From Command% to CommandText%: moved the table and the help and syntax texts
for the commands to just before the SWDL_Code% routine and 'cleaned up' the
texts a bit, but that qualifies as cosmetic.
- Syslog_Name%: not required. Removed.
- Syslog_Version%: not required. Removed.
- From Init% to FinalCode%: several changes, the most important one being the
removal of the claiming of a 16K block of temporary workspace which was
neither used nor released.
There's also the stacking of r7 to r9/r11. It's true that these registers
should be saved (according to the PRM) but only if they're modified. So why
bother if they're not? Replacing this with just stacking r14 enabled a
'cleaner' jump out of the routines on errors.
At the very beginning of Init%, I inserted a check for a value in r12 on
entry. This has to do with re-initialisation after a 'Tidy' operation. Not
sure if that's still required though (hasn't the 'Tidy' functionality been
blocked or removed from RISC OS these days or is that an unfounded rumour?).
I removed everything connected with the CRC-check, including the error text.
SocketWatch is the only module I've ever come across using something like
this and I honestly fail to see what use it is here. But if you do see that,
then by all means leave it in.
In Final% the (unused) ModuleDying% flag was set, but as that had been
removed anyway, so were the three instructions involved in this action.
Actually, as the variable Flags% was re-used to point at the module flags
word, this flag bit isn't set in the first word of the workspace at all. In
fact, on finalisation SocketWatch 0.04y sets bit 1 of the word 592 bytes
from the address r12 happens to point at. That *could* be in the unused 16K
block that was claimed on initialisation, but there's no guarantee. Nasty
(and no, I hadn't realised this yet when I wrote the comment on Flags% in
the Source_004y file...).
FinalCode% (renamed release%) was more or less integrated with Init% (sort
of 'glued to the bottom') but otherwise essentially unchanged, apart from
removing the call to ReleaseBlock% (including the stacking/unstacking of
r14) in favour of a simple MOV r0,#7 / SWI XOS_Module sequence).
- SWI_Register% contained some code at the start (up to the Bvs instruction) I
couldn't make sense of. If you can, please enlighten me. I asked Dickon
about this in my first email to him and he didn't understand it anymore
either. So I removed it.
The rest is essentially unchanged (cosmetics only), but I did replace the
call to GetBlock% and the 'push' and 'pop' instructions around it with the
MOV/SWI instructions that were actually required here.
- SWI_Deregister%: no changes except for replacing the call to ReleaseBlock%
and its surrounding 'push'/'pop' with the necessary MOV and SWI.
- SWI_AtomicReset%: optimisations only.
- SWI_AllocPW%: no changes.
- SWI_DeallocPW%: no changes.
- SWDL_Code%: cosmetic changes only.
02-12-2002 0.04y
-----------------
By Dickon Hood. See the !ReadMe file.
|
