Home - Waterfall Grid T-Grid Console Builders Recent Builds Buildslaves Changesources - JSON API - About

Console View

Legend:   Passed Failed Warnings Failed Again Running Exception Offline No data

esainane
Coerce to number less often in *SetRulesParam (#511)

Unlike most lua_is* functions, lua_isnumber checks to see if the
given value *can* be a number, rather than if it currently is one:

> Returns 1 if the value at the given acceptable index is a number or
> a string convertible to a number, and 0 otherwise.

This means that even if a lua caller expressly coerces to string
beforehand, as long as the string looks like a number, it gets stored
as a number. This can cause problems if, for example, a player name
that looks like a very large number loses precision.

See esainane/zkstats@99ddb35 for an example.

Explicitly checking using the lua_israwnumber variant allows the caller
to provide either a string or a number, and have it be stored as such.
Beherith
Fix 6340 https://springrts.com/mantis/view.php?id=6340, any change in visibity forces subsequence patches Computevariance to be called again.

Patch::ComputeVariance only needs to be called if patch is visible and dirty. Before this change it was called even if tesselMesh was true, as the return value of |= is equal to tesselMesh.
Thus if tesselMesh was set to true previously by a visibility change, then all subsequence patches would get their variance recomputed (which is quite expensive)
Beherith
spaces
Beherith
Port ROAM computevariance fix to maintainance.
abma
fix #6364:

try to add a more generic windows driver crash detection
GoogleFrog
Added the modrule decloakRequiresLineOfSight.
This controls whether sight distance is used as a filter on proximity cloak checks, allowing for cheap implementations of the behaviour desired by FLOZi for s44 and myself for ZK.
Added a missing comment for alwaysVisibleOverridesCloaked.

See:
* https://springrts.com/mantis/view.php?id=6306
* https://github.com/spring/spring/commit/ca1557cdee3fe4075ff9d987dfe0926a229146d6
* https://springrts.com/mantis/view.php?id=6374
Gajo Petrovic
Merge pull request #498 from GoogleFrog/decloakLosOption

Added the modrule decloakRequiresLineOfSight
abma
Merge pull request #504 from Beherith/patch-5

Fix ROAM bug 6340 https://springrts.com/mantis/view.php?id=6340, any change in…
rtri
cp e2d4a1f3c532c19fd87f71cbb8606b0b17dd1893
rtri
close #491
rlcevg
[AI] Fix Command_getParams (#487)

* [AI] Fix LegacyCpp Command_getParams
abma
update CircuitAI
  • full-windows-x64: built develop / default config failed -  stdio
  • linux-static-x64: built develop / default config failed -  stdio
abma
update pr-downloader:
use https for rapid
rlcevg
[AI] Fix Command_getParams (#487)

* [AI] Fix LegacyCpp Command_getParams
Beherith
syntax
abma
update pr-downloader: better error handling / fix exit code not set on error
Beherith
LOAM (Lazy ROAM) for develop branch (#508)

* LOAM (Lazy ROAM) for develop branch

ROAM frees up the entire triangle pool that stores the triangles of each tesselated patch (129*129 hmap pixels) and retesselates all patches whenever any in the visibility of patches occures, or if the unsynced heightmap changes (e.g.) explosion
LOAM does nothing when patches only go out of view, or when patches go into view, and a not-too-stale tesselation is aviable
when a 'stale' patch comes into view, only it is retesselated, and retesselation can affect neighbours, which is tracked and updated accordingly
when a heightmap change occurs, only that patch is retesselated, and if the neighbours are affected, that is updated accordingly
The cost: the mildly increased memory cost of keeping a larger triangle pool updated
the advantage is lag-free scrolling on a map
If the LOAM triangle pool runs out for any reason, then all the stored patches are all reset, and the whole view retesselated (exactly like it happens with ROAM on every visibility change)

* Update LOAM

Added detail increase on zoom in, fixed some midpos errors, shadow meshes are rendered from playercam perspective (much nicer, lower tri load)
Fixed a bug that caused VBO updates for every patch every time

* LOAM less lazy, but distribute load evenly over frames, behave well on zooming :)

* use _STPF_

* type pedancy
  • full-windows-test: created develop / default installer failed -  stdio
abma
update pr-downloader: better error handling / fix exit code not set on error
Jose Luis Cercos-Pita
Fixed invalid collideeMD usage when modInfo.allowSepAxisCollisionTest = true
Beherith
LOAM: cherry pick from develop missed the updates in these two functions (#512)

Causeing reupload of all VBO's every frame, as these functions dont exist in develop
Gajo Petrovic
Merge pull request #496 from Yaribz/playerAdvantageFix

fix SetAdvantage function to allow negative values
Yaribz
fix SetAdvantage function to allow negative values
Jose Luis Cercos-Pita
Fixed invalid collideeMD usage when modInfo.allowSepAxisCollisionTest = true
Gajo Petrovic
Merge pull request #505 from Beherith/patch-6

Port ROAM computevariance fix to maintainance.
Gajo Petrovic
Merge pull request #502 from sanguinariojoe/develop

Fixed invalid collideeMD usage when modInfo.allowSepAxisCollisionTest = true
Jose Luis Cercos-Pita
Stupid brace typo
GoogleFrog
Added the modrule decloakRequiresLineOfSight.
This controls whether sight distance is used as a filter on proximity cloak checks, allowing for cheap implementations of the behaviour desired by FLOZi for s44 and myself for ZK.
Added a missing comment for alwaysVisibleOverridesCloaked.

See:
* https://springrts.com/mantis/view.php?id=6306
* https://github.com/spring/spring/commit/ca1557cdee3fe4075ff9d987dfe0926a229146d6
* https://springrts.com/mantis/view.php?id=6374
Beherith
Fix for RETESSELLATE_MODE == 2 too
abma
hide a lot of compile warnings in assimp
  • full-windows-test: built develop / default config failed -  stdio
  • full-windows-x64: built develop / default config failed -  stdio
Gajo Petrovic
Merge pull request #499 from GoogleFrog/decloakLosOption_maintenance

decloakRequiresLineOfSight for maintenance
abma
update pr-downloader:
use https for rapid
sprunk
Rifle weapontype: remove hardcoded particles (#489)
rtri
cp cd29c09678ef8edf83b28bb27f2c776c5ce6ddde
r-a-sattarov
Added initial support for MCST Elbrus 2000 (e2k) CPU architecture + x86-64 fix (#506)

* x86-64: fixed -march option

* E2K: added initial support for MCST Elbrus 2000

- added initial support in spring engine
- added initial support in 7z lib
- added initial support in asio lib
- added initial support in assimp lib
  • full-windows-x64: built develop / default config failed -  stdio
  • linux-static-x64: built develop / default config failed -  stdio
abma
fix #6364:

try to add a more generic windows driver crash detection
  • full-windows-test: built develop / default config failed -  stdio
  • full-windows-x64: built develop / default config failed -  stdio
Yaribz
fix SetAdvantage function to allow negative values
Beherith
LOAM (Lazy ROAM) for develop branch (#508)

* LOAM (Lazy ROAM) for develop branch

ROAM frees up the entire triangle pool that stores the triangles of each tesselated patch (129*129 hmap pixels) and retesselates all patches whenever any in the visibility of patches occures, or if the unsynced heightmap changes (e.g.) explosion
LOAM does nothing when patches only go out of view, or when patches go into view, and a not-too-stale tesselation is aviable
when a 'stale' patch comes into view, only it is retesselated, and retesselation can affect neighbours, which is tracked and updated accordingly
when a heightmap change occurs, only that patch is retesselated, and if the neighbours are affected, that is updated accordingly
The cost: the mildly increased memory cost of keeping a larger triangle pool updated
the advantage is lag-free scrolling on a map
If the LOAM triangle pool runs out for any reason, then all the stored patches are all reset, and the whole view retesselated (exactly like it happens with ROAM on every visibility change)

* Update LOAM

Added detail increase on zoom in, fixed some midpos errors, shadow meshes are rendered from playercam perspective (much nicer, lower tri load)
Fixed a bug that caused VBO updates for every patch every time

* LOAM less lazy, but distribute load evenly over frames, behave well on zooming :)

* use _STPF_

* type pedancy
Jose Luis Cercos-Pita
Gajops suggested change
Gajo Petrovic
Merge pull request #501 from sanguinariojoe/SATcrashes

Fixed invalid collideeMD usage when modInfo.allowSepAxisCollisionTest = true (maintenance)
Beherith
LOAM (Lazy ROAM) for develop branch (#508)

* LOAM (Lazy ROAM) for develop branch

ROAM frees up the entire triangle pool that stores the triangles of each tesselated patch (129*129 hmap pixels) and retesselates all patches whenever any in the visibility of patches occures, or if the unsynced heightmap changes (e.g.) explosion
LOAM does nothing when patches only go out of view, or when patches go into view, and a not-too-stale tesselation is aviable
when a 'stale' patch comes into view, only it is retesselated, and retesselation can affect neighbours, which is tracked and updated accordingly
when a heightmap change occurs, only that patch is retesselated, and if the neighbours are affected, that is updated accordingly
The cost: the mildly increased memory cost of keeping a larger triangle pool updated
the advantage is lag-free scrolling on a map
If the LOAM triangle pool runs out for any reason, then all the stored patches are all reset, and the whole view retesselated (exactly like it happens with ROAM on every visibility change)

* Update LOAM

Added detail increase on zoom in, fixed some midpos errors, shadow meshes are rendered from playercam perspective (much nicer, lower tri load)
Fixed a bug that caused VBO updates for every patch every time

* LOAM less lazy, but distribute load evenly over frames, behave well on zooming :)

* use _STPF_

* type pedancy