[Hendricks266]

Fox, on 03 July 2012 - 02:59 PM, said:

I wish you could change the start position in Mapster easier than Build. Is it possible to make it so that the start position is displayed as an APLAYER sprite in 2D and 3D mode, and that you could change the position, height or attributes the same way you do with any other sprite?

I think being able to manipulate it in 2D mode is possible and a great idea, including its own little menu to set the exact value of x, y, z, and ang. I don't think 3D mode is a possibility, because...

Fox, on 03 July 2012 - 02:59 PM, said:

I guess it would have an ID of -1, since the sprite doesn't really exist but is just a representation.

It is just about never possible for something to have an ID of -1, whether it is a sprite or a level number. ID numbers are not just a convenience for referencing specific instances of an entity. They are actually indices for arrays, static memory structures which start at 0 and end at MAXTHINGS-1. Using negative numbers not only won't work, it will produce "out-of-bounds" errors which lead directly to crashes. Just ask Helix about those.

EDIT: My thinking is that you treat the start point as its own class of data (which it actually is) alongside sprites, walls, sectors, etc, so you don't bump into the sprite array. This would be very conducive to 2D mode but I don't know what it would entail to display something in 3D mode that is not a sprite yet still have sprite-like qualities of motion.

[Helixhorned]

Fox, on 03 July 2012 - 02:59 PM, said:

I wish you could change the start position in Mapster easier than Build. Is it possible to make it so that the start position is displayed as an APLAYER sprite in 2D and 3D mode, and that you could change the position, height or attributes the same way you do with any other sprite? I guess it would have an ID of -1, since the sprite doesn't really exist but is just a representation.

startpox{x,y,z} and startsectnum are available from m32script. Just insert a dummy sprite, tweak its position to your liking, and set the starting position from the console:

seti <YOUR_DUMMY_SPRITE_INDEX>, set startposx .x, set startposy .y, set startposz .z, set startsector .sectnum, set startang .ang

script_expertmode needs to be enabled.

edit: fixed typo in code.

[Fox]

Hendricks266, on 03 July 2012 - 09:20 PM, said:

I think being able to manipulate it in 2D mode is possible and a great idea, including its own little menu to set the exact value of x, y, z, and ang. I don't think 3D mode is a possibility, because...


It is just about never possible for something to have an ID of -1, whether it is a sprite or a level number. ID numbers are not just a convenience for referencing specific instances of an entity. They are actually indices for arrays, static memory structures which start at 0 and end at MAXTHINGS-1. Using negative numbers not only won't work, it will produce "out-of-bounds" errors which lead directly to crashes. Just ask Helix about those.

EDIT: My thinking is that you treat the start point as its own class of data (which it actually is) alongside sprites, walls, sectors, etc, so you don't bump into the sprite array. This would be very conducive to 2D mode but I don't know what it would entail to display something in 3D mode that is not a sprite yet still have sprite-like qualities of motion.

My point is that it is not a sprite, but is just displayed as one.

[LeoD]

There are newer Ogg/Vorbis versions available, libogg 1.3.0 and libvorbis 1.3.3, which should be put into sources/jaudiolib/third-party/mingw32.

Spoiler

libogg Version 1.3.0 (2011 August 4)

* Add ogg_stream_flush_fill() call
This produces longer packets on flush, similar to
what ogg_stream_pageout_fill() does for single pages.
* Windows build fixes

Version 1.2.2 (2010 December 07)

* Build fix (types correction) for Mac OS X
* Update win32 project files to Visual Studio 2008
* ogg_stream_pageout_fill documentation fix

Version 1.2.1 (2010 November 01)

* Various build updates (see SVN)
* Add ogg_stream_pageout_fill() to API to allow applications
greater explicit flexibility in page sizing.
* Documentation updates including multiplexing description,
terminology and API (incl. ogg_packet_clear(),
ogg_stream_pageout_fill())
* Correct possible buffer overwrite in stream encoding on 32 bit
when a single packet exceed 250MB.
* Correct read-buffer overrun [without side effects] under
similar circumstances.
* Update unit testing to work properly with new page spill
heuristic.

Spoiler

libvorbis 1.3.3 (2012-02-03) -- "Xiph.Org libVorbis I 20120203 (Omnipresent)"

* vorbis: additional proofing against invalid/malicious
streams in decode (see SVN for details).
* vorbis: fix a memory leak in vorbis_commentheader_out().
* updates, corrections and clarifications in the Vorbis I specification
document
* win32: fixed project configuration which referenced two CRT versions
in output binaries.
* build warning fixes

libvorbis 1.3.2 (2010-11-01) -- "Xiph.Org libVorbis I 20101101 (Schaufenugget)"

* vorbis: additional proofing against invalid/malicious
streams in floor, residue, and bos/eos packet trimming
code (see SVN for details).
* vorbis: Added programming documentation tree for the
low-level calls
* vorbisfile: Correct handling of serial numbers array
element [0] on non-seekable streams
* vorbisenc: Back out an [old] AoTuV HF weighting that was
first enabled in 1.3.0; there are a few samples where I
really don't like the effect it causes.
* vorbis: return correct timestamp for granule positions
with high bit set.
* vorbisfile: the [undocumented] half-rate decode api made no
attempt to keep the pcm offset tracking consistent in seeks.
Fix and add a testing mode to seeking_example.c to torture
test seeking in halfrate mode. Also remove requirement that
halfrate mode only work with seekable files.
* vorbisfile: Fix a chaining bug in raw_seeks where seeking
out of the current link would fail due to not
reinitializing the decode machinery.
* vorbisfile: improve seeking strategy. Reduces the
necessary number of seek callbacks in an open or seek
operation by well over 2/3.

Attached File(s)

•  jaudiolib.zip (144.78K)
  Number of downloads: 530

[TerminX]

Cool, thanks. They were never updated because we were too lazy to build new copies I guess. Did you build these yourself or download them somewhere?

[LeoD]

TerminX, on 06 July 2012 - 10:00 AM, said:

Cool, thanks. They were never updated because we were too lazy to build new copies I guess. Did you build these yourself or download them somewhere?

Built myself. MinGW gcc 4.7

[TerminX]

Alright, committed the newer versions. Thanks!

[Helixhorned]

I recently completed the classic renderer's capability of drawing walls with non-power-of-two y sizes:



However, the original maps and imaginably custom ones too already relied on the particular panning with the incorrect drawing. Because the GL modes which came later always draw textures correctly, they had to implement various hacks to make theses cases display like in classic. The result is that now, all three renderers behave differently when faced with such problematic walls.

Here's what I consider correct behavior: the texture wraps once over the ypanning range of 0..255:
Animated PNG 1

This is the legacy classic behavior for reference. The texture is repeated ~2.5 times.
Animated PNG 2

Now things bet hairy. Polymost implemented some "corrections" to be able to display e.g. the "biker bimbos" panel in front of the steroids secret in E1L1, or the picture near the vault in E3L2. However, these are of course really hacks, since it's not easily possible (nor desirable) to reproduce exactly the classic renderer's way of drawing such walls.
Animated PNG 3

Polymer does the panning correction too, but in a different fashion than Polymost.
Animated PNG 4

The reason I'm saying this all is that it is currently a horrible mess. What I intend to do is
1) in the short term, make Polymer's offset "correction" behave like Polymost's. This seems like a sensible thing to do and would fix the picture in E3L2.
2) sometime in the future, new maps will be saved and loaded with the panning values doing what they're supposed to do.

[Fox]

Thank you for that. That's something I was really looking for.

Helixhorned, on 08 July 2012 - 02:48 PM, said:

However, the original maps and imaginably custom ones too already relied on the particular panning with the incorrect drawing. Because the GL modes which came later always draw textures correctly, they had to implement various hacks to make theses cases display like in classic. The result is that now, all three renderers behave differently when faced with such problematic walls.

I don't mind, bug is bug.

[Hendricks266]

Helixhorned, on 08 July 2012 - 02:48 PM, said:

Polymost implemented some "corrections" to be able to display e.g. the "biker bimbos" panel in front of the steroids secret in E1L1, or the picture near the vault in E3L2. However, these are of course really hacks, since it's not easily possible (nor desirable) to reproduce exactly the classic renderer's way of drawing such walls.

I think two more examples are the San Andreas fault sign in E1L5 and the cinema movie on the projector itself (not the screen) in E1L1.

[LeoD]

TerminX, on 06 July 2012 - 10:41 AM, said:

Alright, committed the newer versions. Thanks!

This has brought up some new warnings (yet one left on MinGW):

source/music.c: In function 'MUSIC_ErrorString':
source/music.c:84:21: warning: assignment discards 'const' qualifier from pointer target type [enabled by default]

[Hendricks266]

No, that was caused by r2796, which broke building entirely, then Plagman fixed the build in r2798 but only as a stopgap.

[Fox]

TerminX, on 17 June 2012 - 08:39 AM, said:

I've been spending my limited free time on Max Payne 3 lately (I haven't had time to actually play a game in ages). Making the tile sizes available is a really simple change though... it's literally 2 lines copied from Mapster32 into EDuke32.

Huh, has this been implemented? I failed in finding anything about it in the Changelog and etc.

[fgsfds]

Is this somehow related to this? http://i.imgur.com/WgBz1.jpg
I think this only happens in polymer with top-oriented textures.

[LeoD]

The space(use) key has ceased to work when trying to reload a map after dying.

Attached thumbnail(s)

• 

[Hendricks266]

LeoD, on 09 July 2012 - 07:22 AM, said:

The space(use) key has ceased to work when trying to reload a map after dying.

This is on purpose. r2802:

Quote

Remove BUTTON(gamefunc_Jump) and KB_KeyPressed(sc_Space) from AdvanceTrigger so you can type spaces in savegame names.

You can press Enter, KPEnter, LMB, your "Open" key, and your "Fire" key to advance prompts.

[LeoD]

Hendricks266, on 09 July 2012 - 10:07 AM, said:

This is on purpose. r2802:
You can press Enter, KPEnter, LMB, your "Open" key, and your "Fire" key to advance prompts.

I'm pretty sure that not so long ago I could have spaces in savegames and could reload the map using space...well, maybe some restructuring has happened which now prevents having both.

[Helixhorned]

Fox, on 08 July 2012 - 03:05 PM, said:

Thank you for that. That's something I was really looking for.

That's great, but I have reverted the correct drawing for now because it's the logical thing to do currently: the original levels rely on how those walls are drawn, and we have the correction hacks in place for the GL renderers. So I hope you don't mind waiting until the new map format arrives (this is something that is discussed on IRC from time to time). Levels saved with that one will then always have the correct panning, across all renderers.

LeoD, on 08 July 2012 - 03:29 PM, said:

This has brought up some new warnings (yet one left on MinGW):

source/music.c: In function 'MUSIC_ErrorString':
source/music.c:84:21: warning: assignment discards 'const' qualifier from pointer target type [enabled by default]

Oh hey, I wasn't even aware we had Windows-exclusive code other that winlayer.c and the like up till now. D'oh to that.

Fox, on 08 July 2012 - 07:51 PM, said:

Huh, has this been implemented? I failed in finding anything about it in the Changelog and etc.

No, and I'm not sure it's only "2 lines copied from Mapster32", since as far as I can see CON has no support for system (i.e. non-user) arrays.

empy, on 08 July 2012 - 09:26 PM, said:

Is this somehow related to this? http://i.imgur.com/WgBz1.jpg
I think this only happens in polymer with top-oriented textures.

Absolutely related, though not as easy, and with r2824 they should be identical for Polymost and Polymer.

LeoD, on 09 July 2012 - 07:22 AM, said:

The space(use) key has ceased to work when trying to reload a map after dying.

What troubles me more is that the use key (which for me isn't space) has stopped working as a means to fast-forward the level ending screen. IMO one shouldn't assume a particular binding of keys to functions.

[TerminX]

Yeah, I was in fact wrong when I said it was 2 lines. It actually requires a fair bit more of the array system changes from Mapster32 to be propagated back into EDuke32 itself, which is why it hasn't been done yet.

[Fox]

So I am having a little trouble with Duke Nukem 64 maps. Apparently the geniuses tought it would be a good idea to completely shift the sound numbers. So the maps have a completely different sounds, including many hard-coded sounds.

I have already made a comparison list beetween the sounds in the ROM and the PC. For example, the speech at the beggining of Hollywood Holocaust now uses what originally was the Protozoid Slimer death sound. Sound 396 which is the last sound avaiable is now 255.

How exactly I solve this issue? You can't really make it work in CON because of the hard-coded sounds, and frankly I don't expect to make a workaround with just EVENT_SOUND, since it would take forever and it would never be 100%.

A possibility I tought was of using a Mapster script in these maps. But I don't want to save a map file in Mapster since it will change the size of some sprites and attempt to repair my sectors among other things. And as far I know there isn't an option which if I saved a map in Mapster without actually doing anything every single byte will remain the same.

[Hendricks266]

Fox, on 09 July 2012 - 12:06 PM, said:

A possibility I tought was of using a Mapster script in these maps. But I don't want to save a map file in Mapster since it will change the size of some sprites and attempt to repair my sectors among other things. And as far I know there isn't an option which if I saved a map in Mapster without actually doing anything every single byte will remain the same.

If you can work out preserving the map data, I wrote M32scripts to do exactly what you need, for Vaca+ and NW+. The original code is in the SVN but it's a little bit specialized because I needed to make two passes, at least for picnum switching.

[Fox]

By the way, I have defined a level for E1L8 and there is virtually no way in telling the difference beetween it and an User Map.

Adding a variable that returns whenever you are in a User Map or not may help, but it would only work 100% if I could change the slot occupied by User Map.

[Helixhorned]

Helixhorned, on 09 July 2012 - 11:14 AM, said:

What troubles me more is that the use key (which for me isn't space) has stopped working as a means to fast-forward the level ending screen. IMO one shouldn't assume a particular binding of keys to functions.

By the way, I can't reproduce this any more, and I'm not finding a difference between r2801 and r2802 for my particular case either. Go figure...

[Micky C]

Running r2834, textures such as tile 889 running water no longer animates on walls. They still work on floors and masked walls though. Actually they seem to work on white walls (void space boundary) too. I haven't looked into it much, hopefully it's one of those bugs which is obvious to fix. It's kind of a biggie as it affects the original levels too.

[Helixhorned]

Oops, my bad. Should be fixed with r2835.

[LeoD]

New nedmalloc version available:

Quote

v1.10 beta 3 17th July 2012:

[master 5f26c1a] Due to a bug introduced in sha 7a9dd5c (17th April 2010), nedmalloc has never allocated more than a single mspace when using the system pool. This effectively had disabled concurrency for any allocation > THREADCACHEMAX (8Kb) which no doubt made nedmalloc v1.10 betas 1 and 2 appear no faster than system allocators. My thanks to the eagle eyes of Gavin Lambert for spotting this.

[Mark]

Thanks. For us non-technical types, do we gain something noticable from it?

[LeoD]

Marked, on 19 July 2012 - 10:35 AM, said:

Thanks. For us non-technical types, do we gain something noticable from it?

If you're on a modern OS: most likely not.
WinXP/Linux 2.4 : maybe.
I'm not a technical type in terms of C programming but I wouldn't expect the game to run any faster except maybe for loading times and avoiding loading-related jerks if your PC is at its limits.

[LeoD]

Typing DNSHOWMAP always gives the message: "SHOW MAP: ON", never "OFF".


EDIT - Debian Wheezy 64bit compiler warnings:

source/sdlmusic.c: In function ‘MUSIC_Init’:
source/sdlmusic.c:157:9: warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]

src/driver_sdl.c: In function ‘fillData’:
src/driver_sdl.c:101:8: warning: pointer of type ‘void *’ used in arithmetic [-Wpointer-arith]

src/sdlayer.c: In function ‘getvalidmodes’:
src/sdlayer.c:957:5: warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]


make LTO=1 :

engine.o (symbol from plugin): warning: memset used with constant zero length parameter; this could be due to transposed parameters

obj/xdelta3.o (symbol from plugin): warning: memset used with constant zero length parameter; this could be due to transposed parameters

[Fox]

How is the plan to control the animations before the level starts (like The Birth episode) or after you finish (like when you kill a boss) going in?

Right now I am forced to use some ugly hack that essentially freezes the game by setting all statnums to zero, unless I don't know a better way to do so?