Resman stability and dynamic .jrl file

[elven]

[leo_x]

Sorry, I meant the changes to resman. If the hardcoded directory was cut out it will not be backward compatible.
_________________
the awakening (PW Action)

[elven]

[elven]

Alright, one more thing -

Apparently, nwnx_physfs doesn't reload script files for me in a physfs dir. I just copy them in, but any ExecScript in nwserver still runs the old code. Is this because of the ncs cache? Still haven't gotten my head around that. :(

[leo_x]

Yea, sorry... the physfs was whipped up quick to test the resman events. So it only returns something once and will never invalidate the cache. I'll work on fixing all this tonight. As to the physfs directories, I might have to expose the write directory. Since physfs is for games it might expect everything to be static excepting the write directory.
_________________
the awakening (PW Action)

[leo_x]

I uploaded an update. I renamed the version of resman i was working on to nwnx_resources and have updated physfs to use it's new event code.

See here: https://github.com/jd28/nwnx2-linux/tree/ta/plugins/resources
and here: https://github.com/jd28/nwnx2-linux/tree/ta/plugins/physfs

There is no more resman dir. I only tested basic functionality. Physfs was able to serve script files from a directory, no memory leak, and no crashes. I wasn't able to test hot patching files, since I don't have a module on this computer. I exposed the PhysFS write director via nwnx2.ini setting 'write_dir', but I'm not sure that it will be necessary.

I will test more when I get home and have access to a module.
_________________
the awakening (PW Action)

[elven]

Thanks for your quick reply! I've just tested your new code.

With the same setup (just load0 = /path in nwnx2.ini) and resources + physfs, no script reloading happens if file mtime changes. Not sure what the "physfs write directory" is - I'll ask google later.

Exists doesn't seem to be called - I sprinkled it with some log entries to try and debug this and they never appeared. Then I've added debug logs to DemandRes and it looks looks like PHYSFS_getLastModTime(resref) always returns 0.

[leo_x]

In the phsyfs library it looks like an getLastModTime should resolve to stat on posix systems... and should return -1 on error. But then it does the same if a zip file is mounted too... I'll have to take a closer look. The next version is supposed to have a stat style function.

Hmm, that is odd. I don't see anything obviously wrong with the event hook stuff. It's no printing an error that it wasn't able to connect. It's interesting, I wonder if resource existance check just doesn't get called for all types...and nwserver just bails on a NULL resource data. I see some GIT's are being tested but no NCS or 2da. I never looked closely enought at resman to see notice if it did the same. If that's the case then I think the architecture will have to change again to do some generic query event before any demand...
_________________
the awakening (PW Action)

[elven]

That's a bugger. Not sure what to do about this - I have no clue about resman internals and a very big list of things that needs doing before digging into that some more.

Let me know if I can help you test stuff, at least (and you can catch me on irc - irc.nwn2source.net/#nwn2cr - too if you want).

[leo_x]

Just looked through the dissamebly and it looks like CExoResMan::Exists is called in very few places. CopyItemAndModify, LoadGIT, some DM messages, and module load... so it is important that it exists in the plugin, but it's not called every Demand unless it got inlined everywhere else...

Thanks, I'll keep working on it.

I've been meaning to give the IRC a try. When are people most active?
_________________
the awakening (PW Action)

[elven]

They're not. :) It's more kind of a connect-and-lurk thing. Activity happens all the time, but sporadically - there's about 10-20 people in, give or take.

But it's a good place to lurk, set some highlight notifications and catch people if something comes up.

[leo_x]

I'll definitely check it out.

I just pushed an update. I was able to get it load updated NCS files... The issue isn't super clear to me but I think some the resref buffers were ill-formed. There will some weird prints in the logs too. Switching to std::string cured all that.

One weird thing remains... The mtime is getting set properly and prints correctly on the nwnx_resource side, but on the nwnx_physfs side it still prints 0. I forget the proper format thing for long long, maybe that's why?

In any case, it worked with both directories and zip files.
_________________
the awakening (PW Action)

[elven]

Found another small bug. Not sure where it's coming from:

I'm trying to load an area via nwnx_areas from physfs. Says this in the log:

[Thu Sep 4 15:06:59 2014] File: anr_n001o001u00.are exists 0

area resref is anr_n001o001u000.are, which is within spec (16 chars).

Can confirm it's not nwnx_areas, which reads:

[Thu Sep 4 15:05:16 2014] Creating area 'anr_n001o001u000'
[Thu Sep 4 15:05:16 2014] Loading area 'anr_n001o001u000'
[Thu Sep 4 15:05:16 2014] Load failed: 'anr_n001o001u000'

Had a quick look at the code in resman (and physfs), but didn't find any issues that would point at that. It's also happening in nwnx_redis or any other provider. You got an idea why it's cutting of the last char?

[elven]

Also, I'm seeing the occasional crash when replacing script resources inside physfs.

Not sure what's up with that either:

[leo_x]

There's something definitely wrong with what's in the repo at this moment. I was getting double frees at shutdown myself. Not had much chance to work on it tho...

As for the truncated character, I did see that issue before. I think it's even in the orignal resman: https://github.com/NWNX/nwnx2-linux/blob/master/plugins/resman/NWNXResMan.cpp#L84 could be wrong about that tho. If not, I probably copied it over. In this new stuff I was just going to say screw it and use std::string.
_________________
the awakening (PW Action)