gamedevlists-general — List for discussing general game development issues

[GD-General] Re: asset & document management

[Enno R. <en...@de...>]

Thatcher Ulrich wrote:

> On May 15, 2003 at 01:54 +0200, Enno Rehling wrote:
> 
> We're using Perforce; it handles binary assets just fine.  We have
> done a bunch of scripting to automate some content processes, e.g. so
> that artists can hit a button in Maya to do the appropriate
> edit/checkout.  This seems to be working pretty smoothly for us.

How does it handle an artist that creates 182 versions of a 80 megabyte 
binary file in the course of 3 weeks? I suppose CVS would end up with 14 GB 
of archives, by which time it has probably long croaked :-)

I've looked at Perforce in the past, it's a really nice product, simliar to 
what we already know. But does it handle really, really large amounts of data?

Enno.

[GD-General] Re: asset & document management

[Enno R. <en...@de...>]

Mickael Pointier wrote:

> So back to our own product. It's minimalistic in the sense that it does not
> perform archiving and cannot merge. Basicaly it's an "Exclusive Checkout"
> based system that simply allow people to get/put/add/checkout/checking files
> from a common network repository.

I've also had a look at unison, because someone on the list mentioned it. I 
don't like the idea of not being in control over the archiving. On the one 
hand, I clearly don't want to keep all the old versions.

If I can't buy the asset management that I like, this is what I think I'd 
like to build:

CVS has a lot of features (webcvs, logs, branching, tags, etc.) that I 
wouldn't want to miss. Also, I really like TortoiseCVS, because it has the 
"integrate in the explorer shell" functionality you mentioned, and that's 
making many people feel a lot more comfortable than a command line, 
especially designers and artists.

On the other hand, large binaries that undergo several changes a day are not 
manageable in CVS in the long run. What I would really want is to be able to 
say "only keep the last 5 versions, treat the rest as if they are 0-length 
files". It might be possible to extend CVS like that - the cvs files are 
fairly easy to read, and I could write an external process that locks a 
directory once a day, and kicks out old versions. Obviously, this only works 
on binaries where I don't have diffs, but always full files.

In addition, I would like to be able to "nail down" an archived version, so 
it doesn't get flushed out. Anthing with a tag on it, for example, would be 
kept.

If that's workable, I'd get the power of CVS, but not the huge storage 
requirement. And something that the artist already know from other places.

Anyone who sees a problem in that?

Enno.

[GD-General] Re: asset & document management

[Enno R. <en...@de...>]

Kent Quirk wrote:

> At 01:54 PM 5/15/2003 +0200, you wrote:
> 
> A good bug tracker is indispensable. Has bugzilla gotten any easier to 
> install since last year? I looked at it and ran away in horror.

Well, I really can't remember, it's been so long. I think you really just do 
apt-get install bugzilla and go through the configuration dialogs. You need 
a bit of understanding as to what it is you're installing there, but nothing 
horrid.

We've since adapted bugzilla a lot, and use it for all task tracking. 
Basically, nobody does anything that isn't in bugzilla, which gives us a lot 
of control and predictability. People do time estimates on their tasks, we 
use milestones extensively, and we're thinking of tying it closer together 
with CVS (in our system you can already get from the webcvs changes to the 
bugzilla task that caused the change, but as of now, you can't do it the 
other way round). And we have some rudimentary export from bugzilla to 
project to make gant charts of what's going on. nice.

> Annoyingly, by far the best product of this type I've ever used is Lotus 
> Notes. I say "annoyingly" because it's got a strange design dating back 
> to 1993. 

I rmember Notes from the mid-nineties, at university. It was weird, and 
while I can see that it would do parts of what we want, it's jsut so clunky 
to use that I can't see it being widely accepted here.

> But with all that said, we're now moving to custom XML designs for all 
> our document types. We've built the beginnings of a general-purpose XML 
> forms editor tool (there are such things available from people like 
> Altova as well). Having the data in an XML format means that it's pretty 
> easy to build/find tools to transform, catalog, index, and search in the 
> files. But you also want a decent editor to use while working with them.

So you make your design documents, slides, project plans, etc. all in those? 
Or are you saying you use the XML for indexing the documents that you make? 
Anyway, we're not big enough to spend a lot of time on building our own 
software solutions, I'm afraid.

> I'd also say that artist's opinions on data organization are highly 
> suspect. It's been my experience (and yes, I'm generalizing) that 
> artists have to be dragged kicking and screaming into any kind of 
> structured data organization. They're visual people and tend not to care 
> that much about silly things like textual names. The key is to find a 
> system that lets you find the assets you need with minimal impact on 
> artist productivity.

Jep, absolutely. OTOH, I've had some that were absolutely delighted after 
they learned about CVS and what it offers them. For small, textual 
resources, that is.

We've pretty much rolled our own in the past, too, and it's just not as 
powerful or hassle-free as a good commercial product would be, so I'd really 
like to switch. :-/

Enno.

Re: [GD-General] asset & document management

[Thatcher U. <tu...@tu...>]

On May 15, 2003 at 01:54 +0200, Enno Rehling wrote:
> 
> The other thing I miss is asset management for large binaries -
> stuff that we can't or shouldn't put into CVS. I know there's
> alienbrain, but for some reason the evaluation didn't leave our
> artists very happy. It's also not exactly cheap, IIRC. Do you know
> anything else, are you using something that makes you say "wouldn't
> know how I could live without it"?

We're using Perforce; it handles binary assets just fine.  We have
done a bunch of scripting to automate some content processes, e.g. so
that artists can hit a button in Maya to do the appropriate
edit/checkout.  This seems to be working pretty smoothly for us.

My (third-hand) understanding of AlienBrain is that they are adding
value by including a lot of this convenience scripting and integrating
for popular tools.

-- 
Thatcher Ulrich
http://tulrich.com

Re: [GD-General] asset & document management

[Mickael P. <mpo...@ed...>]

>> The other thing I miss is asset management for large binaries -
>> stuff that we can't or shouldn't put into CVS. I know there's
>> alienbrain, but for some reason the evaluation didn't leave our
>> artists very happy. It's also not exactly cheap, IIRC. Do you know
>> anything else, are you using something that makes you say "wouldn't
>> know how I could live without it"?
>
> We did large assets ourselves through a custom system that stored
> them on a server and kept a database catalog of them. Alienbrain
> initially looked more like a tool kit than an actual application, but
> I've heard lately that it's gotten better.
>
> I'd also say that artist's opinions on data organization are highly
> suspect. It's been my experience (and yes, I'm generalizing) that
> artists have to be dragged kicking and screaming into any kind of
> structured data organization. They're visual people and tend not to
> care that much about silly things like textual names. The key is to
> find a system that lets you find the assets you need with minimal
> impact on artist productivity.
>
> Seems like a lot of people build their own, here.

Yep, here we also checked Alien Brain, when it was still called "NxN
MediaStation" back in 1999, and at this time it was painfully slow
(something like 4 minutes between the moment when you openned the client
view and get the data ready on screen...), and was more looking like a "Do
it yourself" package. Since we didn't had an unlimited amount of time to
invest in that, we tried an alternative solution: developping our own
minimalistic system.

Since this first version it seems that AlienBrain had been optimised a lot,
but so far I never find any review of the product made by thirdparties on a
real project that does not look like a copy of the white papers available on
their website, so I will wait for someone telling me it worked great for
them on a real product before going into evaluation again...

So back to our own product. It's minimalistic in the sense that it does not
perform archiving and cannot merge. Basicaly it's an "Exclusive Checkout"
based system that simply allow people to get/put/add/checkout/checking files
from a common network repository.

It's totaly made using Windows 2k+ shell extensions, so it's cleanly
integrated in the windows explorer (right click contextual options, addition
columns to display the status of files, property page to show the history of
file, and so on...). Among the options are a small "send message" that
appear when user select a file that someone else has in checkout. Using
windows message service so it allow the user to send a "please could you let
me the access to that file, thanks" that appear as a modal box on the screen
immediately (like network printer message feedback). Also it has a small
"view difference" option that can redirect to any 3rd party differencing
software depending of the file type (pictures using our own picture
difference program, for text types it can be windiff or araxis merge...).
Got the idea ?

Our artists are using it because it didn't force them to use other tools
than before. They can perform checkin/checkout from any explorer instance,
meaning it works also in applications fileselectors. Also, since it does not
perform archiving, the speed is similar to that of a straight copy from/to
the network, so they don't get the illusion of losing time when working.

It's not a great product, but well, it works. Been used on three products
(one beeing finished on 3 different targets) without any major problem.

    Mickael Pointier

Re: [GD-General] asset & document management

[Kent Q. <ken...@co...>]

At 01:54 PM 5/15/2003 +0200, you wrote:
>What kind of tools do other game developers use?
>
>I've been a die-hard CVS fan for many, many years now, and two years ago, 
>added bugzilla to the list of tools I wouldn't want to work without. But I 
>realize that there are still things I lack.

A good bug tracker is indispensable. Has bugzilla gotten any easier to 
install since last year? I looked at it and ran away in horror.

>One thing is document management. A game can create lots and lots of 
>documents, and just putting them into revision control isn't enough - I 
>need a way to search for them, I want fulltext search through all of them, 
>keywords, categories, compare revisions, and more - or at least a 
>significant subset of those features. I worked on a project that used MS 
>Sharepoint for this, and while that was okay, it had its own set of 
>problems (It still became messy after a while, and it didn't work in 
>non-IE browsers, plus it's not good at storing html documents). Has anyone 
>had good experience with other such software?

Annoyingly, by far the best product of this type I've ever used is Lotus 
Notes. I say "annoyingly" because it's got a strange design dating back to 
1993. But it has the best metaphor for managing sets of related documents 
I've ever seen, and in modern versions you can use it pretty effectively 
without having to install the Notes client on everyone's desktop. We have 
used it for everything from bug tracking to design databases to actual code 
repositories and code generators for components that had to fit within a 
special template.

But with all that said, we're now moving to custom XML designs for all our 
document types. We've built the beginnings of a general-purpose XML forms 
editor tool (there are such things available from people like Altova as 
well). Having the data in an XML format means that it's pretty easy to 
build/find tools to transform, catalog, index, and search in the files. But 
you also want a decent editor to use while working with them.


>The other thing I miss is asset management for large binaries - stuff that 
>we can't or shouldn't put into CVS. I know there's alienbrain, but for 
>some reason the evaluation didn't leave our artists very happy. It's also 
>not exactly cheap, IIRC. Do you know anything else, are you using 
>something that makes you say "wouldn't know how I could live without it"?

We did large assets ourselves through a custom system that stored them on a 
server and kept a database catalog of them. Alienbrain initially looked 
more like a tool kit than an actual application, but I've heard lately that 
it's gotten better.

I'd also say that artist's opinions on data organization are highly 
suspect. It's been my experience (and yes, I'm generalizing) that artists 
have to be dragged kicking and screaming into any kind of structured data 
organization. They're visual people and tend not to care that much about 
silly things like textual names. The key is to find a system that lets you 
find the assets you need with minimal impact on artist productivity.

Seems like a lot of people build their own, here.

         Kent


>_______________________________________________
>Gamedevlists-general mailing list
>Gam...@li...
>https://lists.sourceforge.net/lists/listinfo/gamedevlists-general
>Archives:
>http://sourceforge.net/mailarchive/forum.php?forum_id=557

Kent Quirk, CTO, CogniToy
ken...@co...
http://www.cognitoy.com 

[GD-General] asset & document management

[Enno R. <en...@de...>]

What kind of tools do other game developers use?

I've been a die-hard CVS fan for many, many years now, and two years ago, 
added bugzilla to the list of tools I wouldn't want to work without. But I 
realize that there are still things I lack.

One thing is document management. A game can create lots and lots of 
documents, and just putting them into revision control isn't enough - I need 
a way to search for them, I want fulltext search through all of them, 
keywords, categories, compare revisions, and more - or at least a 
significant subset of those features. I worked on a project that used MS 
Sharepoint for this, and while that was okay, it had its own set of problems 
(It still became messy after a while, and it didn't work in non-IE browsers, 
plus it's not good at storing html documents). Has anyone had good 
experience with other such software?

The other thing I miss is asset management for large binaries - stuff that 
we can't or shouldn't put into CVS. I know there's alienbrain, but for some 
reason the evaluation didn't leave our artists very happy. It's also not 
exactly cheap, IIRC. Do you know anything else, are you using something that 
makes you say "wouldn't know how I could live without it"?

Enno.

[GD-General] CHM files

[Pierre T. <p.t...@wa...>]

Hi,

I'd like to create docs as .chm files. Doxygen supports this but I'd like to
create a user-manual for an app, not something for programmers, and I think
Doxygen only works with source files (tell me if I'm wrong here), when I'd
like to work from already existing html files.

So I tried HTML Help Workshop, which works fine and does what I want. But I
have a slight problem with it, and I can't find any info on the web about
this : sometimes, images (jpegs) are not included in the resulting CHM
(though they are correctly displayed in source html files).

Any ideas ?

Thanks,
- Pierre

Re: [GD-General] Re: Hardware Survey

[Parveen K. <pk...@al...>]

On Tue, 2003-04-29 at 04:01, Enno Rehling wrote:
> Daniel Vogel wrote:
> >>>That 40% for Linux servers makes it very clear that we absolutely have
> >>>to do a dedicated linux port, sigh <g>. At least we don't have to
> >>>worry about the client -- 0.69% -- very generous for you guys to even
> >>>bother with a client port, it clearly had no financial benefit.
> >>
> >>That's actually a good question. Daniel, what was the motivation behind
> >>building and supporting a Linux client?
> >=20
> >=20
> > Mark summed it up nicely in his post to our forums.
> >=20
> > http://www.ina-community.com/forums/showthread.php?s=3D&threadid=3D2011=
70
>=20
> I'm a latecomer to this list, and reading archives. This thread has=20
> disappeared from the boards, can anyone point me to a mirror?

IIRC, the conversation went something like:

"We are impatient fanboys.  Why are you wasting time with a Linux
client?  We want the game yesterday."

"We are intelligent developers.  And we realize that porting our engine
to many platforms results in a stronger product."

Of course, I'm paraphrasing a little bit.

--=20
Parveen Kaler <pk...@al...>
http://www.sfu.ca/~pkaler

RE: [GD-General] Re: Hardware Survey

[Daniel V. <vo...@ep...>]

> I'm a latecomer to this list, and reading archives. This thread has
> disappeared from the boards, can anyone point me to a mirror?

The summary basically is that if we have a Linux server there will be a
Linux client.

To make this slightly on-topic again, it's actually much easier to test the
server if you have the client running on the same OS as well as it's not
exactly fun to track down issues that only appear in client/server
scenarios.

In the past we (or to be more precise, Ryan Gordon) usually went down the
road of porting "ucc" (the standalone command line tool) and once that is up
and running you basically have most of the engine running at which point you
can focus on rendering and sound. That usually uncovers the remaining
nastyness if you port to a different endian or 64 bit system.

-- Daniel, Epic Games Inc.

[GD-General] Re: Hardware Survey

[Enno R. <en...@de...>]

Daniel Vogel wrote:
>>>That 40% for Linux servers makes it very clear that we absolutely have
>>>to do a dedicated linux port, sigh <g>. At least we don't have to
>>>worry about the client -- 0.69% -- very generous for you guys to even
>>>bother with a client port, it clearly had no financial benefit.
>>
>>That's actually a good question. Daniel, what was the motivation behind
>>building and supporting a Linux client?
> 
> 
> Mark summed it up nicely in his post to our forums.
> 
> http://www.ina-community.com/forums/showthread.php?s=&threadid=201170

I'm a latecomer to this list, and reading archives. This thread has 
disappeared from the boards, can anyone point me to a mirror?

Enno.

RE: [GD-General] Re: stl on consoles

[Andrew G. <ag...@cl...>]

The debug symbols aren't loaded by the console they're used by the debugger.
In the early days of SN's GC tools we were getting debug executables of
about 150mb!

FWIW we don't use the STL, we rolled our own versions of the most commonly
used parts which have identical signatures, we did the same for large parts
of the standard C library.

Andy


-----Original Message-----
From: Enno Rehling [mailto:en...@de...] 
Sent: 28 April 2003 05:11
To: gam...@li...
Subject: [GD-General] Re: stl on consoles


I can't speak from experience about stl on consoles, haven't used it. But 
one problem I see that you haven't mentioned and I haven't seen discussed is

symbol size. STL debug symbols tend to be big - I wrote a sample application

using luabind the other day, which uses boost and some funky template stuff 
to create C++ bindings for lua. The executable was 400 K without debug 
symbols, and 16 MB with debug symbols. That wouldn't have fit on a console, 
ever.

I think that for XBOX at least, you can keep the debugging symbols off the 
actual console, in a PDB file, but I'm not sure if that's the way all 
consoles handle it - thinking about it, they probably do. Has anyone got 
more console experience than me?

Even then, the excessive inlining you get eith STL code is going to bloat 
your executable, and on some platforms you can't grow it much. On the 
gamecube, Nintendo won't let you make executables bigger than 4MB, for 
example, so this could conceivably become an issue.

Enno.




-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf _______________________________________________
Gamedevlists-general mailing list Gam...@li...
https://lists.sourceforge.net/lists/listinfo/gamedevlists-general
Archives: http://sourceforge.net/mailarchive/forum.php?forum_id=557


DISCLAIMER: Unless otherwise expressly stated, this message does not create
or vary any contractual relationship between you and Climax Studios Ltd.
The contents of this e-mail may be confidential and if you have received it
in error, please delete it from your system, destroy any hard copies and
contact the originator of the email. In accordance with the
Telecommunications (Lawful Business Regulations) (Interception of
Communications) Regulations 2000 the Company reserves the right and, may at
any time, monitor and intercept (but not record) e-mails to establish if
they are relevant to its business.

Re: [GD-General] Re: [mlist] RE: Off the shelf scripting languages

[Thatcher U. <tu...@tu...>]

On Apr 28, 2003 at 02:35 +0200, Enno Rehling wrote:
> Thatcher Ulrich wrote:
> 
> >On Apr 23, 2003 at 12:05 +0300, Ivan-Assen Ivanov wrote:
> >
> 
> >In general, the Lua stack will be chock full of pointers to live
> >objects, as will the Lua objects on the heap.  So as Tom says, you
> >serialize this just like you serialize anything else -- traverse the
> >live objects, keep track of object ID's, fix up pointers, etc.
> 
> I've just started looking at lua and whether that would be a good 
> replacement for our own scripting. I started by looking at coroutines and a 
> simple scheduler and ran across the same issue of serializing the state for 
> savegames. Are there any good examples on how to traverse the lua stack and 
> find my C pointers to do the serialization? I'd be thankful for a few 
> pointers :-)

That in particular should be pretty straightforward; just iterate over
the stack elements (using the Lua API) and look for entries which are
userdata.  But, besides your own data, there are various parts of the
VM that need to be serialized as well, not to mention Lua's heap
objects, which may also reference your objects.  The Lua
garbage-collector code (src/lgc.c) is very similar to what you need
for serialization.

I think the lua-l mailing list would be a better place to ask, though.

-- 
Thatcher Ulrich
http://tulrich.com

[GD-General] Re: A free sound lib able to play MP3 and WAV with pitch control ?

[Enno R. <en...@de...>]

Lionel Fumery wrote:

> To answer to myself, I found a good sound library with pitch control here :
> http://www.nctsoft.com

Speaking of sound libraries, I recenly discovered audiere 
(http://audiere.sf.net/), and it did exactly waht I needed it for, with an 
API that was so simple it made me cry joyful tears :-)

Enno.

Re: [GD-General] A free sound lib able to play MP3 and WAV with pitch control ?

[Lionel F. <li...@mi...>]

To answer to myself, I found a good sound library with pitch control here :
http://www.nctsoft.com

LiF

[GD-General] Re: [mlist] RE: Off the shelf scripting languages

[Enno R. <en...@de...>]

Thatcher Ulrich wrote:

> On Apr 23, 2003 at 12:05 +0300, Ivan-Assen Ivanov wrote:
>

> In general, the Lua stack will be chock full of pointers to live
> objects, as will the Lua objects on the heap.  So as Tom says, you
> serialize this just like you serialize anything else -- traverse the
> live objects, keep track of object ID's, fix up pointers, etc.

I've just started looking at lua and whether that would be a good 
replacement for our own scripting. I started by looking at coroutines and a 
simple scheduler and ran across the same issue of serializing the state for 
savegames. Are there any good examples on how to traverse the lua stack and 
find my C pointers to do the serialization? I'd be thankful for a few 
pointers :-)

Enno.

[GD-General] Re: stl on consoles

[Enno R. <en...@de...>]

I can't speak from experience about stl on consoles, haven't used it. But 
one problem I see that you haven't mentioned and I haven't seen discussed is 
symbol size. STL debug symbols tend to be big - I wrote a sample application 
using luabind the other day, which uses boost and some funky template stuff 
to create C++ bindings for lua. The executable was 400 K without debug 
symbols, and 16 MB with debug symbols. That wouldn't have fit on a console, 
ever.

I think that for XBOX at least, you can keep the debugging symbols off the 
actual console, in a PDB file, but I'm not sure if that's the way all 
consoles handle it - thinking about it, they probably do. Has anyone got 
more console experience than me?

Even then, the excessive inlining you get eith STL code is going to bloat 
your executable, and on some platforms you can't grow it much. On the 
gamecube, Nintendo won't let you make executables bigger than 4MB, for 
example, so this could conceivably become an issue.

Enno.

Re: [GD-General] A free sound lib able to play MP3 and WAV with pitch control ?

[Nalin S. <ns...@vs...>]

try AMP11Lib and OpenAL.
AMP11LIB can play MP3 files quite well and it was used by Croteam for their
game "Serious Sam" and is the basis for the Serious Engine's sound rendering
capabilities.

OpenAL is used by Garagegames in their award winning Torque engine (which is
the basis of the Tribes2 game) and is an opensource 3d sound rendering
library. Learn more abt it at http://www.openal.org

Hope that helps...

Regards,
Nalin

This lib with
----- Original Message -----
From: "Lionel Fumery" <li...@mi...>
To: <gam...@li...>
Sent: Saturday, April 26, 2003 4:09 AM
Subject: [GD-General] A free sound lib able to play MP3 and WAV with pitch
control ?


> Hello,
>
> Does anybody know where I could find a good free lib that can playback MP3
> and WAV with a pitch control ? (yes, I need the pitch control in a video
> game :))
> With or without DirectSound, doesn't matter for now... (evaluation step !)
>
> Thank you in advance for your comments,
>
> LiF
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Gamedevlists-general mailing list
> Gam...@li...
> https://lists.sourceforge.net/lists/listinfo/gamedevlists-general
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_id=557

[GD-General] A free sound lib able to play MP3 and WAV with pitch control ?

[Lionel F. <li...@mi...>]

Hello,

Does anybody know where I could find a good free lib that can playback MP3
and WAV with a pitch control ? (yes, I need the pitch control in a video
game :))
With or without DirectSound, doesn't matter for now... (evaluation step !)

Thank you in advance for your comments,

LiF

Re: [GD-General] Convert 3ds to quake3 bsp

[gekido <ge...@ub...>]

there is a .map exporter from milkshape3d i believe.

should be able to open it with q3radiant or whatever afterwards

mike w
www.uber-geek.ca

Frank Puig wrote:
> Hi
>   I need to convert a 3ds (or any file that the 3dsmax can export) to a 
> quake3 bsp file. Could any one can give me a hint. I have the q3radiant 
> but It don't load 3ds and it save to .map. But I need a .bsp and the 
> "bspc"don't give me information about how to compile .map to .bsp
>  
> What I'm missing?
>     Frank
>  
> 
> 
> ------------------------------------------------------------------------
> Do you Yahoo!?
> The New Yahoo! Search 
> <http://us.rd.yahoo.com/search/mailsig/*http://search.yahoo.com> - 
> Faster. Easier. Bingo.

[GD-General] Convert 3ds to quake3 bsp

[Frank P. <fpu...@ya...>]

Hi  I need to convert a 3ds (or any file that the 3dsmax can export) to a quake3 bsp file. Could any one can give me a hint. I have the q3radiant but It don't load 3ds and it save to .map. But I need a .bsp and the "bspc"don't give me information about how to compile .map to .bsp What I'm missing?    Frank 


---------------------------------
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.

Re: [mlist] RE: [GD-General] Off the shelf scripting languages

[Thatcher U. <tu...@tu...>]

On Apr 23, 2003 at 12:05 +0300, Ivan-Assen Ivanov wrote:
> 
> To support serialization, the interpreter stacks must not only
> NOT be intertwined with the C stack - the "stackless" property of
> Stackless Python, and, I presume Lua 5.0 - but also must not contain
> any live pointers to real memory - or provide a mechanism for
> serializing the state of each execution thread.

In general, the Lua stack will be chock full of pointers to live
objects, as will the Lua objects on the heap.  So as Tom says, you
serialize this just like you serialize anything else -- traverse the
live objects, keep track of object ID's, fix up pointers, etc.

-- 
Thatcher Ulrich
http://tulrich.com

RE: [mlist] RE: [GD-General] Off the shelf scripting languages

[Tom F. <to...@mu...>]

We don't have "threads" in our engine - just state. And every game turn,
code operates on the state to produce new state. Of course the code that
happens is mainly controlled by the state, but we don't have persistent
thread state. So it's not a problem.

We've shied away from multithreading stuff like this. I know it's cool and
trendy, but it's also completely evil to debug, nasty to save (as you say),
and we're pretty used to doing stuff this way.


Tom Forsyth - Muckyfoot bloke and Microsoft MVP.

This email is the product of your deranged imagination,
and does not in any way imply existence of the author.

> -----Original Message-----
> From: Ivan-Assen Ivanov [mailto:as...@ha...]
> Sent: 23 April 2003 12:55
> To: gam...@li...
> Subject: RE: [mlist] RE: [GD-General] Off the shelf scripting 
> languages
> 
> 
> > Python has "pickling" to do this. Or you can do it the way 
> > you do everything else - assign all objects a unique ID, and 
> > save the stuff out manually, replacing pointers with IDs. 
> > Which is essentially what pickling does.
> 
> Pickling is serialization of Python objects from within Python code;
> what I'm talking about is pickling the thread state itself.
> 
> Pickling thread state in Stackless is announced as "possible in
> the future" and "coming really soon" by Tismer, the author of 
> Stackless,
> on several occasions.
> 
> What do you do for serializing the scripting environment state?

RE: [mlist] RE: [GD-General] Off the shelf scripting languages

[Ivan-Assen I. <as...@ha...>]

> The latest version of Lua (5.0) is mostly "stackless", so 
> latent functions are supported, and so you don't have 
> troubles with interpreter state embedded in the C stack (if 
> that's the problem you're referring to).  You would still 
> have to write some code to serialize the VM and heap state.

There are two separate problems.

To support latent functions, the interpreter must be able to yield
(as in "yield vs. return" when you talk about coroutines)
control back to the thread scheduler that runs it when it encounters
a latent function call - this is more convoluted than the usual
"run this script for it's side effects and return a value when it's
done"
model of embedding.

To support serialization, the interpreter stacks must not only
NOT be intertwined with the C stack - the "stackless" property of
Stackless Python, and, I presume Lua 5.0 - but also must not contain
any live pointers to real memory - or provide a mechanism for
serializing the state of each execution thread.

regards,
Assen

RE: [mlist] RE: [GD-General] Off the shelf scripting languages

[Ivan-Assen I. <as...@ha...>]

> Python has "pickling" to do this. Or you can do it the way 
> you do everything else - assign all objects a unique ID, and 
> save the stuff out manually, replacing pointers with IDs. 
> Which is essentially what pickling does.

Pickling is serialization of Python objects from within Python code;
what I'm talking about is pickling the thread state itself.

Pickling thread state in Stackless is announced as "possible in
the future" and "coming really soon" by Tismer, the author of Stackless,
on several occasions.

What do you do for serializing the scripting environment state?