Listing all open sockets

I’m thinking that the address should come before the port, not after it.

I’m also thinking that the all-zero IPv6 address should be represented as [anyhost], the same as for IPv4 sockets. That means that there would be no way to tell whether a listening socket is IPv4 or IPv6, which means adding another column.

I’m thinking that the address should come before the port, not after it.

I’d agree, it’s the way I normally think about things when working.

there would be no way to tell whether a listening socket is IPv4 or IPv6, which means adding another column.

I don’t think that’s true. To quote a chunk of your earlier post:

Active Internet connections (including servers)
Proto   Recv-Q Send-Q  Local Address          Foreign Address        (state)
ip           0      0  *.49171                *.*                   
ip           0      0  *.32771                *.*                   
ip           0      0  *.137                  *.*                   
ip           0      0  *.32770                *.*                   
Active Internet connections (including servers)
Proto   Recv-Q Send-Q  Local Address          Foreign Address        (state)
ip6          0      0  *.443                  *.*                   
ip6          0      0  *.80                   *.* 

ip vs. ip6

You didn’t suppress DNS lookup

I didn’t mean addresses by name, I meant it didn’t give me lots of zeroes.

All zeros for a local and remote address means an IPv4 listening socket for the given local tcp or udp port number.
For IPv6 the syntax is ::port_number for listening ports.

Updated screenshot of work in progress:

No guarantee that this picture will remain available – if I delete it, I’ll have to try to remember to delete or edit this post.

ResRec causes Zeropain logs to be generated. That’s going to be fun to find.

Updated screenshot of work in progress:

Port after IP is more logical, more space required between IP and port perhaps? Or may a dot or colon ?

ResRec causes Zeropain logs to be generated.

Does the original version do that? If not, then it’s time to concentrate on the changed items, like the use of the newer Socket_ calls you will have needed to bring in the IPv6 info

That’s going to be fun to find.

Well, there’s zero assembler in the program, so logically it has to be something it calls

That’s going to be fun to find.

Well, there’s zero assembler in the program, so logically it has to be something it calls

You do not need assembler in BASIC to risk ZeroPain errors – the usual cause is not something it calls, but indirect accesses to memory … and there are hundreds of those in ResRec!

I will try and dig out my notes about coping with the ‘fun’ with BASIC ZeroPains.

Does the original version do that?

Yes, and it doesn’t appear to be related to sockets – it also happens when I double-click a zip file on a local drive.

The result is 128 ZeroPain logs, time stamped to the same second. I can see R0 incrementing from 7 to &86, and R2 incrementing from &8100 to &817F.

Martin: thanks for ZPdebug, it is proving useful. The errant lines caused by opening files and sockets are very similar. Each contains multiple statements, so the next step is to split them into lines each containing a single statement.

The code there does stuff I don’t yet understand, but anyone who wants to look at the original code should look for DEFPROCscanfile and DEFPROCscansocket.

There’s a condition:

IF F% > PAGE AND $(F% + 8) <> A$ THEN

It’s sometimes entered with F% = -1.

The same condition appears in two functions, and a very similar condition with the same error occurs a bit lower down in one of them. I think I’ve fixed them all now.

It’s sometimes entered with F% = -1.

Just discovered that here as well … in PROCscanfile. But I have also seen F%=0.

There is stuff that I do not understand either :-((
Waiting 10cs between checks of individual file handles seems excessive to me!

IF F% > PAGE AND $(F% + 8) <> A$ THEN

That’s one of the downsides to BASIC… In C (or, indeed, Python and many other languages), that would be a valid way to avoid the NULL dereference, but in BASIC you’d have to do something like

IF F% > PAGE THEN IF $(F% + 8) <> A$ THEN

as in BASIC, AND and OR are bitwise operators and not boolean ones that will short-circuit.

Personally, I find the BASIC approach a lot less readable, but appreciate that it’s a matter of taste.

Next significant problem is to improve the window redraw – it doesn’t shrink when the number of lines reduces, and lines can remain there after they’ve closed.

Steve: I’ve put the conditions into multi-line functions.

Hmm – Using a pointer as a (yes no) info?

It’s sometimes entered with F% = -1.

Just discovered that here as well … in PROCscanfile. But I have also seen F%=0.

Dumb question, but is there any reason it’s even allowed to enter when F% is less than or equal to zero?

Or is this one of those “we ought to check passed parameters, but it works, error free, without the check” (until it doesn’t) situations?

Dumb question, but is there any reason it’s even allowed to enter when F% is less than or equal to zero?

There is an ELSE too. Clearly, when F% is -1, the IF clause isn’t executed (but the ELSE is), but the interpreter still has to evaluate the rest of the logical expression before deciding the overall condition is FALSE.

Martin: your ZPdebug has found why my modified !Printers RunImage file causes ZeroPain errors on startup. It’s a very similar mistake that I wrote into my logging code, which attempts to log an indirected string without first checking that the pointer is valid. Quickly fixed. Thanks again!

Re. the code that does stuff we don’t understand: I think it’s stuff that could be done with arrays, but the author chose to do it in his own way, managing memory himself. Remember, the original ran in a 32 kiB Wimp slot. Mine doesn’t any more, simply because I’ve undone the compression of the bits I’ve been working on, so as to make them clearer to read. I chose to give it a 64 kiB slot, which should more than account for my fattening.

Makes Gerfs point of testing small functions the best you can :-)

Makes Gerph’s point

There, fixed that for you. ;)

but the author chose to do it in his own way, managing memory himself

Perhaps because the data is liable to change a lot and BASIC doesn’t have dynamic arrays?

I chose to give it a 64 kiB slot

Argh! No! That’s double!
<looks at TaskManager saying 933064K free>
Umm, as you were then…

which should more than account for my fattening.

If I remember correctly, HIMEM - END will tell you how much space is left over.

HIMEM – END will tell you how much space is left over.

Not quite, as that includes the stack. If you use *ReportMem Addr it gives the whole story, eg…

Memory: Prog=16,373 Vars=8,592 Free=3,920 Stack=40 Undefined=0 Slot=32768 
PAGE=&8F00 TOP=&CEF5 LOMEM=&CEF8 END=&F088 r13=&FFD8 HIMEM=&10000 MEMLIMIT=&10000

Ref TV Quest Repair Shop – doing a repair of a Mechanical calculator – times past :-)

I’ve put version 0.45 on my web site: https://davehigton.me.uk/

Andrew Pullan’s version does not appear to have any copyright or licence statement anywhere, so I believe it’s acceptable for me to release this version. It is unlikely to work on versions of RISC OS prior to 3.7 – if you really are going to continue with such an ancient version of RISC OS, it’s best to stick with Andrew’s version.

As usual, please let me know how you get on with it.