Lilly Zenovka took my source changes for the old school version 18i and compiled them for Linux. She gave me a few changes to the source in areas where my source caused hiccups in gcc and added Linux code to display memory footprint. These changes will go into my future releases also, so anybody who wants to make Linux compiles will have it a bit easier.
There's also a downloadable binary for Linux, so if you want to run Lilly's compiled Linux version, point your web browser to her server: http://emak.free.fr/src/sl-nicholaz-18i.tar.gz
Thanks Lilly :-)
Friday, August 24, 2007
Subscribe to:
Post Comments (Atom)
15 comments:
thank you both.
18xf on its way :)
18xf for linux. A dream will come true :)
Thank you Lilly, looking forward to get my hands on it :)
Heyas, I started to work on the Mac build of 18i, too. There were a few minor nitpicks (a multiline comment and a missing return in one file, one - not needed - header include for a header that doesn't exist on OS X) and the built went fine. It created a monstrous thing, though - guess there are tons of symbols in it or something like that, but I guess that's a missing setting in the mac XCode project.
But two weird bugs did show up:
- the cursor doesn't change when you use the alt-cam feature (keeping alt pressed and moving the mouse to move your cam)
- the map is broken. The low-resource jpegs are shown fine, but when they snap into focus (get sharp and viewable), they suddenly revert to 1/4 of the actual map part - only the upper left part of the map is shown.
Any ideas on what that might be? Very strange ...
Barney,
I don't have an idea about the mouse cursor, but the map thing sounds a bit as if my sanity check in llimagej2c.cpp might get in the way (grep for "EXPERIMENTAL" in the llimage folder ... the place is pretty obvious).
Also, have you tried a clean (non nicholaz) compile to see if there's something else that could cause this?
Haven't yet tried a clean build, as I first wanted to get a look at the nb build process and get the stoppers out of the way. I am currently running the universal build to see what that produces (if it again produces >700 MB executables, I am crying!) and will check the clean build next.
Thx for the tip on the map problem, will look into that one, too (as soon as this build is done - takes about 30 minutes to do a full build, have no idea what the universal build - targetting two platforms - will take).
There seems to be the problem that the open source build for the mac doesn't produce the same results as the delivered binary from LL - I stumbled over that a bit ago allready. For example the wrong icon is set and some other thigns are different, as if they don't actually use the same xcode project file for their deployment ...
Ok, I found the giant-executable-problem (as exepcted it was project settings in the build file) and tried your hint with the experimental stuff in the llimagej2c.c - I changed the #IF to 0 and disabled your code. But after rebuilding, the same results. The map is out of focus and as soon as it is in-focus, only the top left corner of each region on the map is in focus, the rest is again out of focus (as if a step before the last is shown there) and the in-focus part is wrong-colored.
I cleared the cache to be on the safe side with the map images, not that they are leftovers from a previous run. Same results. Any more ideas where that might happen? I'll start a clean build next, to be sure it's not something in the original code from LL.
Any hints on where in the code to dig for the cursor-not-showing-during-alt-cam problem, or this map-not-showing-right problem? I haven't yet fully internalized the source structure, so would be grateful for directions ;)
Wow !!
cant hardly wait to give this a try
Kool ~ thanks
JayR Cela
Since an image says more than hundred words, here the borked map:
http://radio-boomslang.shacknet.nu/~bb/images/borked-map-1.jpg
http://radio-boomslang.shacknet.nu/~bb/images/borked-map-2.jpg
(image 2 shows it best)
Barney,
if you look at llimageworker.cpp and diff it with the original (or look at nicholaz\patches\0000_imagestuff.patch) there's something that may affect this too (the change with new LLImageRaw around line 100). I think you can simply put the normal llimageworker.cpp back in place to test this.
Hmm. the 0000_imageworker.patch is missing from your 18i download file - will have to look at the source tree itself (it's in the linden/ subtree, but not in the patches/ subtree). Other changes seem to be missing from the patches, too - for example I couldn't find the patch responsible for removing the version check.
But I guess it's all moot anyway: my clean build shows the very same problems. I havbe one last chance, that is an update to XCode 2.4, maybe some GCC or XCode bugs bite me here - but I am not too optimistic (I am still running XCode 2.3).
Yes, with the 18i version there's a bit of a gap between the source and patches (especially in areas that are purely nicholaz specific). It's a bit better with 18xf because I need to carry those changes over to the next Linden release, but 18i is more like a forking than a branch.
Ok, I throw in the towel - it's not worth the hassle. If I download a software source and can't reliably reproduce the distributed binary from that, it's pointless to even try to do any compiles. I built the binary with the 2.4.1 xcode now and have the same stupid errors with the broken map and the cursor changes missing. LL doesn#t seem even to be able to produce a useful source package, at least it definitely can't be the source package and library package they use to build the official client - the project settings are wrong and there are some bugs in it that aren't in the released version. Bah. that sucks.
Oh dang, what a pointless waste of time! That so tremendously sucks!
The only idea I have would be to ask on the sldev mailing list if anyone else encountered it already.
Lilly is indeed awesome ;)
Post a Comment