What changed in EDuke32, build by build

This fixes the vertical (~1 texel wide) stripes that are especially noticable with tiled art skies and Polymost with r_parallaxskyclamp=1 or Polymer (which always uploads skies with repeating s and t coords). What constitutes a sky tile is currently hardcoded to be >=78 and <=99.

was only used there, from the source.

and restore_highlighted_sector functions. What this means is that TROR bunches are now correctly duplicated too (only if all sectors of a given bunch are highlighted).

(i.e. first press RAlt, then hold RShift to make it active), consider all sectors, even the grayed out ones. This makes 6 modifier keys in 4 categories for RAlt. Can you name them all? Also spell out the labels that get drawn near the reticle when applying some of the modifiers and make them not show when not actually selecting (e.g. pressed RShift alone, but have highlighted sectors)

that bunches and nextwall links are restores, too. Now, to coalesce two TROR maps (e.g. for a CBP), one could do for i allsectors, sethighlightsector i 1 on the console, and then load the 'other' map to copy over the contents of the first, including all TROR stuff. Note that neither selecting all sectors with RAlt, nor duplicating them with Insert would lead to the desired effect (yet). Internally, - yax_setbunch() now has different behavior for bunchnums -1, -2 and -3 (it either clears the nextwalls or not) - build.c has a helper function free_n_ptrs(). This could be handy in other situations where many blocks of memory must be allocated consecutively and freed in case of failure

suboptimal, but at least 'correct' and won't produce the aforementioned corruptions anymore.

nextwalls. (i.e. setting them to -1 on 'corruptcheck tryfix')

can be caused by copying map parts from one map to another (or a new one). The copying system is slightly broken right now....

next{wall,sector} indices instead of using checksectorpointer() so that the nextwalls will be restored exactly as they were (provided that they are copied). Also fix tweaking the sectnums of sprites.

Called 'loop punching'. In addition to duplicating the loop, the inner portions of the original and cloned loop are made into a new bunch, and the loop walls are made to be neighbors, linking their movement to each other. This gives a way to create island sectors with a differing bunchnum in the midst of an already extended area. Usage: For an extended sector containing a CCW inner loop, select its points using Ctrl+Shift. Press Ctrl-E to be queried whether to duplicate the points into the upper or lower extension (even if it's unambiguous). A number of conditions must be met for a successful 'punch', which should hopefully be more or less obvious. Among them: - loop walls should all be red - the starting wall is only considered among the non-grayed-out ones - the neighboring sector must not contain points inside or at the loop

would break out of the 2D mode loop when splitting a sector would exceed limits.

- Don't crash when sector joining fails under certain circumstances. - Increase the point lock (Manhattan) distance from 1/16th to 1/8th of the grid square length, making it easier to snap to vertices not lying at the grid points with grid lock on - When drawing walls, don't snap to any of them except the first drawn

memory map.

(lazy_tileselector=1 in mapster32.cfg). We're drawing tiles one by one, so glDrawBuffer(GL_FRONT_AND_BACK) is our friend.

sector c/f and moving one of them would go beyond the other side (ceilingz > floorz), don't move any sector c/f at all. When moving only one sector in that case, move so that ceilingz == floorz.

Now, if more than one sectors are affected, always move them by the same amount. That is, if one of them would be sticking against the other side, don't move the rest. Another change consists of determining the z delta first, and only then moving the sprites by that amount. This fixes the problem where you'd e.g. raise a 'floor door' against the ceiling, and the effector sprites would be raised one time too much, preventing them from being lowered the next time.

- new mapster32.cfg and OSD variable 'fixmaponsave_sprites', telling the editor whether to 'fix' the sprite sectnums on saving and entering 3D mode. Note that a) sprite sectnums are never fixed when 'script_expertmode' is enabled, as before, and b) sprites that have out-of-bounds sectnums are still fixed if a proper sector is found - if saving and some sprite sectnums have thus been tweaked, inform the mapper on the status line and print the changes in the OSD - fix update issue similar to an earlier one: when 'saving as', the astub.c- based file name is now updated too, so saving with Ctrl-S now saves into the one save with 'save as' (and not the old one) - don't attempt to recheck wall pointers from scratch (when saving and running with -check) if it's a TROR map

This was meant for debugging only.

Hopefully this fixes the periodic pausing/unpausing that a few people have been experiencing. Also fix build with GCC 3.

the 'optimize' function attribute if LTO is enabled.

with RAND_MAX > 32767 (everything except Windows?). The 'displayrand' command now returns values from 0 to 32767 inclusive, across all platforms.

setactorsoundpitch <actor#> <sound#> <pitchoffset> which can be used to change the pitch of a playing sound. The pitch offset has the same meaning as the definesound pitch range endpoints, i.e. the units are 1/100th of a seminote. Note that just like the random pitch offset, increasing the pitch makes the sound duration shorter (and vice versa).

member, so maybe not *that* bad) and an invalid memory access when trying to access a per-player gamevar in a player-less context (e.g. WGR2 SVN r25 WGR2GAME.CON line 10448).

playback functions. Without these, there would be an uninitialized access in line 444 of multivoc.c (if the memory in question was allocated with malloc, but still...)

and some VoiceNode struct members with a const, commit forgotten vorbis.c changes.

MV_MaxVolume into one again.

last one + this one: A couple of tweaks against accidental overwriting of maps. - first, fix a bug where the map name wasn't reset to 'newboard.map' when starting a new map - When saving with Ctrl-S, query the user for confirmation and also show the file name where the map will be written - when going into the ESC-menu, print the file name where the map will be written on pressing 'S' above the '(S)ave' text

- first, fix a bug where

Keep in mind that MUSICANDSFXs with ambient sounds must still have lotags less than 999 because values >=1000 are used for the amount of reverb. (999 can't be used because the original code reads '< 999' and I'm not sure whether it has any special significance... probably not though)

cursor, query the mapper to pick one. The joining can still fail afterwards, which makes it only a semi-automatic helper feature.

allowed circle-walls started on the unconstrained _other_ side of a constrained wall. Also fix a potential invalid wall access when inserting a point (pretty serious, that!)

also for make-circle (C, SPACE). There's still a TROR corruption happening probably when inserting on constrained walls, but I've yet to get behind the real cause.

some wall drawing state is ripped out from overheadeditor() into a a file-scope struct (even if we don't yet need it there). If the deleting touches the first drawn point, the walls drawn so far are cleared.

ATI cards. From the time of OpenGL initialization, gamma lookup in classic/fullscreen will be done in software. Before that, always do a setgamma() after setting the palette, since this is what resets the HW gamma. However, this may lead to unappealing artifacts with portions of the scene being sent to the screen with different gammas.

add it to the 'all' target instead source/midi.c: eliminate one 'variable set but not used' warning

hack.

maybe replace them with the proper names. Usage: checkdefs.sh <some.def> [[<some_dir>] -patch] <some_dir> is taken to be the base directory of the search path. -patch uses 'sed -i' to replace the offending findings

warnings with GCC 4.6 except one in kplib.c

fashion. Among other things, this makes the FOV widening when shrunk work. User CON code should always check the initial viewingrange value when attempting to do FOV effects like zoom, because it is not guaranteed to be 65536 due to different screen aspect ratios etc.

First, when pressing 'S on signed members (like the shade), display the real signed value instead of the value cast to an unsigned type. Second, when aiming at an overwall and pressing G, affect the overwall instead of the wall. Maybe these changes affect a few other queries too...

ceilings or floors. In such cases, you're not squished unconditionally (instead of fetching the ceiling or floor height of the neighbor sector), but TROR isn't meant for small-scale RoR anyway.

polymer_addlight(). I have no idea whether it's the 'right' thing, but it fixes this: Conditional jump or move depends on uninitialised value(s) at polymer_updatelights (polymer.c:5090) by polymer_drawrooms (polymer.c:938) . . . Uninitialised value was created by a stack allocation at loadmaphack (engine.c:9482) Fix another uninitialized access when rendering jpegs.

This fixes a crash when loading a map afterwards. Tweak a few debugging messages and some whitespace.