On Fri, Jun 13, 2014 at 03:32:26PM +0100, Bob Latham wrote: Couple people try moving it elsewhere (ie, outside !Boot.Resources), and configure !Boot to "look at" !Cache *BEFORE* it looks at or runs !NetSurf, and see if this solves the problem? I'm wondering if there's something odd about memory availabilty at certain points at boot, or if the !Run/!Boot of !Cache needs tweaking. B.

On Fri, Jun 13, 2014 at 09:42:54PM +0100, Andrew Pinder wrote: Thanks, that's a useful data point. We'll look more closely. B.

Andrew Pinder, on 14 Jun, wrote: !Cache does not error here when in !Boot.Resources, but it also cannot use its SysLog calls as !SysLog is booted after !Cache. If I try to use !Boot's 'Look at' the error here is "Not enough memory", and the cache doesn't. One work around is to comment out the SysLog run line in !Cache.!Run. Another work around is to Filer_Run <SysLog$Dir>.!Run which had the disadvantage the the SysLog module is not loaded in time for !Cache.!RunImage to use it and is therefore somewhat useless. A more useful evasion is to ensure the the SysLog module is loaded before !Cache is booted. That is what is happening here but so far I have not seen the "Not enough application memory to start Basic" error. Raspberry Pi, OS5.21 (03-Jun-14). Hope this helps. -- David Pitt

On 14/06/14 13:48, David Pitt wrote: Indeed. The original proposal for Cache by Rob Kendrick[1] simply set two variables in its !Run file. These were: Caches$Write - Expands to where you should write new cache data Caches$Path - Expands to where you should read cached data We recently decided to make use of "!Cache" as we assumed it would be suitable and correct. (None of the core developers are actually using RISC OS at the moment, so we hadn't checked it actually worked.) Looking at the content of the supplied "!Cache" it seems to contain a lot of extra stuff and unexpected complexity. As shipped, it currently looks like this: !Cache/ !Cache/!Boot,feb !Cache/!Help,feb !Cache/!Run,feb !Cache/!RunImage,ffb !Cache/!Sprites22,ff9 !Cache/!Sprites,ff9 !Cache/Caches !Cache/Caches/Blank !Cache/Caches/Single !Cache/Caches/Single/MimeMan !Cache/Caches/Single/MimeMan/MimeBase,ffd !Cache/Caches/Single/MimeMan/PluginMime,ffd !Cache/Resources !Cache/Resources/MultiError,ffb !Cache/Resources/ResFind,ffb !Cache/Resources/UK !Cache/Resources/UK/Help !Cache/Resources/UK/Messages !Cache/Resources/UK/!Meta !Cache/Resources/UK/Templates,fec While we'd expect something like: !Cache/ !Cache/!Boot,feb !Cache/!Help,feb !Cache/!Run,feb !Cache/!Sprites,ff9 !Cache/!Sprites22,ff9 !Cache/Caches We can either create our own "!Caches" that behaves as we expect, or we can try to get the upstream "!Cache" fixed and continue to use that. Although "!Cache" does seem over-complex; I'm not sure why it needs a !RunImage, for example. Also, whichever we choose, we should move the RUfl cache to the persistent cache, rather than storing it in Scrap. [1] See this thread: http://vlists.pepperfish.net/pipermail/netsurf-users-netsurf-browser.org/... -- Michael Drake http://www.netsurf-browser.org/

On 14/06/14 14:30, Andrew Pinder wrote: Thanks, but I think it's already been tested enough to show it doesn't work, as it stands. You should be able to put it in !Boot without it throwing up errors or having other side-effects. Our options are to get Cache from http://www.snowstone.org.uk/riscos/ fixed, and continue using it, or create our own simpler solution. Steve Fryatt: any thoughts on this? Cheers, -- Michael Drake http://www.netsurf-browser.org/

On 14 Jun, Michael Drake wrote in message <539C509C.8090109(a)netsurf-browser.org&gt;: Well, I can't reproduce the problems that people are seeing. That's on RISC OS 5, with reasonably recent copies of !Boot, and !Cache placed in !Boot.Resources then left to be processed from there. I don't run NetSurf during the boot sequence -- are all the people seeing issues doing that? What are people's Next slots configured to during the boot process? The contents of !Cache don't seem too unreasonable. The !RunImage is just there to set some system variables in a sensible way and make double-clicks on the application behave differently once it's been set up (to open the cache folder). As Chris says, there's Select profile support, and it logs its setup decisions to SysLog if that's available. You could probably do the whole lot from within an obey file, but it would get messy. I'm not sure what the MimeMan cache stuff is, but wonder if that got left in by accident. It looks to be cache data for MimeMan. -- Steve Fryatt - Leeds, England http://www.stevefryatt.org.uk/

On 14 Jun 2014 Steve Fryatt <lists(a)stevefryatt.org.uk&gt; wrote: I don't see any problems either, FWIW. Don't run NetSurf at Boot, and Next at Boot is set at 1024K. ARMini, RISC OS 5.19 at the moment, with !Cache in Resources. No idea of any of this is useful info. [snip] Best wishes, Peter. -- Peter Young (zfc W) and family Prestbury, Cheltenham, Glos. GL52, England http://pnyoung.orpheusweb.co.uk pnyoung(a)ormail.co.uk

In message <d55ad31754.Andrew-Pin(a)waitrose.com&gt; on 14 Jun 2014 Andrew Pinder <Andrew.Pinder(a)tiscali.co.uk&gt; wrote: Adding WimpSlot -min 16k -max 16k to the !Run file just before Run <Cache$AppDir>.!RunImage gets rid of the error message. It does, however, lead the application to think it has not been invoked via its !Boot file, so it opens the Cache directory. This is tested via SYS "XOS_ReadVarVal","Cache$FromBoot",block%,-1 TO ,,exists% Cache$From!Boot is set in the !Boot file before the !Run file is called and then unset immediately afterwards. Regards Andrew -- Andrew Pinder

In article <9ed6811754.Andrew-Pin(a)waitrose.com&gt;, Andrew Pinder <Andrew.Pinder(a)tiscali.co.uk&gt; wrote: I'm using ARMiniX 5.21. I've tried hiding !cache so I can control when it is filer_booted but it still gives the problem. Bob.