D4R issues

Looks like version 0.01a was a botch!
It has got a hardcoded IP for my internal server, not sure how that got missed – but apologies for that!

It has got a hardcoded IP for my internal server,

You too huh? ;-)

The way I do this now is to embed a magic name that is used for debugging, in my case something like http://azumi/blahblah...

Edit: just remembered if I use “azumi.home” the router will DNS that to the PC that’s running the server. Or “raspberrypi.home” for the Pi, “kiseki.home” should be the iPad (not sure this one works), and “akirei.home” ought to be this phone on the network unless it does something dumb like identify itself as “samsung.home” or the like.

I’ve just eaten, but a quick google search turned up this
Really hungry again now!

OK I’ve uploaded version 0.02a – essentially the same except I’m not setting the default value as my internal test server!
One thing to note is that it ONLY supports 200 requests, my last test was to download the fedora iso.
The link is a 302 to the nearest mirror, so all that happens is it saves the headers to the file.
I’m not sure why it saves the headers to the file, I don’t think it should…something else to look at!
I hadn’t even thought about dealing with 302 etc!
But anyway I downloaded the iso without issues and the sizes seem to match so it all looks good.
Time to crack on with getting the initial dodgy GUI interface done…..
But you were right about being brave using OS_Lib, time to put in a for loop to trap all event handlers again from 0 to 999999 so I can work out the event number for finished drag.
It really shouldn’t be this painful every time! The documentation is nearly worthless. (well maybe not if you know the whole code base, but then whats the point?)
Sometimes I wonder, I’m almost considering finding the title bar name of the window under the pointer rather than working out how drag and drop works….
I did this years ago when I was a teenager, sheesh my brain has shrunk!

I think I’m having a really bad night.
I have a function I call when the download button gets called.
inside this function I have:
if (( e = button_set_value( 0, Download_WindowHandle, 4, “what??” ) ) != NULL )
{
wimp_report_error(e,0,0);
exit(1);
}

The Download_WindowHandle is being set correctly, the icon number is 4 and nothing happens and there is no error trapped.
The button_set_value is definately being called, as if I put an exit before or after it, my program exits.

My thoughts are that it should work and there is possibly a buffer overflow being caused somewhere resulting in some “undefined” behaviour?

A long time ago, I wrote a couple of simple C applications using OSLib and OSLibSupport (for Toolbox). I thought it was pretty good. From what I remember, OSLib maps all the SWIs to easy-to-use C functions. So it’s not so much the OSLib Help that will tell what you need… more likely the regular SWI documentation in various StrongHelps and the PRMs.

I’m using toolbox for most of my stuff, not OSLib

Malcolm, did you fix your problem yet? Can’t see the issue myself atm. Have you tried doing a button_get_value on the same window/component to see if you get what you’d expect back? Might shed some light…

Just on the subject of OSLib, I definitely recommend you switch from RISC_OSLib and the Acorn toolbox library to OSLib. IMHO, it’s much, much better.

Just on the subject of OSLib, I definitely recommend you switch from RISC_OSLib and the Acorn toolbox library to OSLib. IMHO, it’s much, much better.

Yes. OSLib has very comprehensive support for the Toolbox (the official Toolbox – there isn’t any support for third party releases such as Tabs, but then I assume the Acorn toolbox library is the same).

Malcolm: What are you using? You complained about OSLib, but then said ‘Toolbox’. I’m confused.

You’re confused, that makes two of us ;-)
I’m using toolbox

I have given your second version a quick test. Tried a zip and a jpeg from my own (remote) web site. It did not error in the same way as before. However, in both cases it just hourglassed for a long time (mins). It did create an empty file on the hard drive, ‘apparently’ 16384 bytes long. The only way out in the end was Alt-break, when the file reverted to zero length. The activity lights on the ethernet switch showed very little activity, so I do not think any downloading occurred at all.

Netsurf downloaded the same zip in a couple of seconds.

Not much help I am afraid.

Malcolm: What are you using? You complained about OSLib, but then said ‘Toolbox’. I’m confused.

I think Malcolm is using the C stubs provided as part of the DDE, and referenced in the Toolbox manual itself. I don’t know if it’s classed as part of RISC_OSLib or not, but I expect that’s what he’s using. OSLib’s interface is of course almost identical, but there are subtle differences, and IMHO, OSLib’s overall coverage is more consistent and logical.

I think there may be some confusion here between the different library names.

RISC_OSlib is different to OS_Lib – the former is was Acorn’s first attempt at a “slightly higher level” WIMP library (amongst other things), whilst the latter provides friendly C veneers to almost all RISC OS swis with sensible types/structs etc defined. It’s “close to the metal” but makes for friendlier code.

As such, it’s entirely practical to code toolbox apps using OSlib as a veneer to swis (including the toolbox swis themselves) using as much or as little of Acorn’s Toolbox support libraries as you desire.

mmmm.
Hourglass issue, that is most bizarre.
That section is using the same codebase as newsuk, as there are no hourglass parts in the http code.

This is an example of my “testing”

*SDFS::HardDisc0.$.Programming.!D4R.D4R download.fedoraproject.org /pub/fedora/linux/releases/21/Workstation/x86_64/iso/Fedora-Live-Workstation-x86_64-21-5.iso
SDFS::HardDisc0.$.test/iso
Downloading http://download.fedoraproject.org/pub/fedora/linux/releases/21/Workstation/x86_64/iso/Fedora-Live-Workstation-x86_64-21-5.iso to SDFS::HardDisc0.$.test/iso.
*

This doesn’t really work as it returns headers in the file, which contain the Location: Header like so:

HTTP/1.1 302 Found
Date: Wed, 11 Feb 2015 19:32:34 GMT
Server: Apache/2.4.6 (Red Hat) mod_wsgi/3.4 Python/2.7.5
Content-Length: 0
Location: http://mirror.sov.uk.goscomb.net/fedora/linux/releases/21/Workstation/x86_64/iso/Fedora-Live-Workstation-x86_64-21-5.iso
AppTime: D=21808
AppServer: mirrorlist-host1plus.fedoraproject.org
Content-Type: text/html; charset=UTF-8
ProxyTime: D=415863
ProxyServer: proxy01.phx2.fedoraproject.org
Connection: close

but if you then change it to the download location in the Location: Header….e.g.

*SDFS::HardDisc0.$.Programming.!D4R.D4R mirror.sov.uk.goscomb.net /fedora/linux/releases/21/Workstation/x86_64/iso/Fedora-Live-Workstation-x86_64-21-5.iso SDFS::HardDisc0.$.test/iso

This works OK for me, so I’m guessing your input values are wrong.
I’m not validating anything at the moment.

The first argument is the dns name
The second is the location of the remote file
The third is where you want to save it
I’ve only used one space to separate the arguments.
You cannot specify a port number.
So www.paymentlabs.com:80 will not work as it will attempt to do a DNS lookup on “www.paymentlabs.com:80”

The hourglass issue, can only be either a DNS issue or a problem connecting to your server.
The current RISC OS resolver code is….
“interesting”?

Given the fact that the link you are using works using netsurf I’m guessing you’re passing something “funny” to D4R.

Could you send me the params you are using and I’ll take a look.
I’m intregued to see what the “wrong” values are, it could be any number of things I haven’t thought about!

malcolm at paymentlabs dot com ;-)

Another thought, my d4r binray is 33504 bytes.
I was wondering if something has a “cached” version?

The current RISC OS resolver code is….
“interesting”?

For a particular definition of “interesting” :)
I haven’t had a moan about the resolver stuff this month. Not feeling well.

My problem there is writing the resolver code looks fairly straight forward, but the module side of things I’m struggling with.
If I wrote a load of functions “gimme_a_record” “gimme_mx_record” (well not that exactly)
I’d be quite happy to BSD (or whatever) it if someone wants to “slot” it in, or help me “slot” it in.

If I wrote a load of functions “gimme_a_record” “gimme_mx_record” (well not that exactly)
I’d be quite happy to BSD (or whatever) it if someone wants to “slot” it in, or help me “slot” it in.

The module shell that Rick put up the other week in reply to a query would probably provide your wrapper.

The module shell that Rick put up the other week in reply to a query would probably provide your wrapper.

Module wrapper → https://www.riscosopen.org/forum/forums/3/topics/3073#posts-39049

Steve’s tweaks → https://www.riscosopen.org/forum/forums/3/topics/3073#posts-39056

HTTP/1.1 302 Found

It isn’t unreasonable for a person to think that the http fetcher would follow the redirect itself (perhaps spitting a message to the user to tell them this).

[from the other thread]

for downloading large files, where you may want to resume or pause the downloads – this was for the WROCC club videos (PLUG PLUG) as they are around 500MB.

Umm… YouTube? I’d expect an entire movie-length video (SD-res) for that sort of file size! Or are you using an ancient codec like MPEG2?

This is getting a bit confusing with two threads being used almost alternately 8)

OK. I’ll give it another go tomorrow. I am pretty sure I was setting up the command line correctly, but I will doubly check next time. The d4r binary is indeed 33504 bytes, so I guess I am using the correct one. I was doing this on the Panda. Maybe I’ll try on different hardware if I still do not get any success.

Update. Have just tried it on the Iyonix and it seems to be working correctly on that. Strange. I’ll check on the Panda again tomorrow.

OK. Gave the Panda a full reboot (normally on 24/7) and things are hunky dory. (Was chasing a nasty little bug a day or so ago which was trashing memory in one of my apps I was developing – must have done more than just in-app damage).

Tried a few zips from remote sites – it is very fast compared with Netsurf as you say, and is probably being limited by my relatively slow connection – only syncing at about 4 Mb/s so max transfer will be stuck at about 330 MB/s (and this is on the edge of Edinburgh!).

I’ll try some batch downloads via an obey file later just for the hell of it.

RISC OS is a nightmare for things like this. Sockets etc. seem to be the worst for trashing things badly.
I managed to trash the NIC driver a few times when I was testing newsuk in the alpha versions.

@Rick 302’s will be handled, the videos are designed to be played on KinoAmp, so are mpeg 1 or 2. The videos are around 1-1.5 hours long.

A base module isn’t my problem, debugging it is. (another abort on data transfer/filecore in use…)
I’ll take a look and try and do some DNS work tonight.
That’s the userland part not the module part, I’m not touching modules for as long as I can (which unfortunately for me, may not be that long)

Steve’s tweaks

Bit of a grand description for a line defining a SWI as a number at the start and a semicolon at the end as a reminder that the “END” needs a following carriage return.

It would be a bit better if the SWI definition was replaced with
GET header.SWIs-all

where a sub-directory called header had a file called SWIS-all which was a list of all the SWIs in the system defined in the same form sort like:

OS_WriteC	EQU	&000000

OS_WriteS	EQU	&000001

OS_Write0	EQU	&000002

OS_NewLine	EQU	&000003

OS_ReadC	EQU	&000004

OS_CLI	        EQU	&000005

OS_Byte	        EQU	&000006

Well the good news is I’ve managed to get a working version of a dns client in around 2 hours.
Now before you think I’m the fastest coder that has ever existed, I was banging my head against the wall reading the header format for DNS etc.
Then I stumbled across an ANSI C compatible dns client, downloaded it and with a minor tweak of the sources (seriously minor) it works.
Plus it’s designed to be async as well!
The only issue is the fact that I haven’t got a clue where riscos stores it’s equivalent of resolv.conf!
So it’s hard coded to 8.8.8.8 ;-)
*tadig news.bbc.co.uk
news.bbc.co.uk: 212.58.246.83

The licence is amusing, but it’s definately worth a beer!
Anyway, if anyone is feeling like a little laugh – head over to the d4r download page and grab it.
No source yet, as I want to modify the example code a lot to make it work properly for RISC OS.
But the library pretty much worked straight away! Impressive!