llGetSubString("abc", 3, -1) gives "abc"; llList2List(L, 1, -1) gives L when L has length 1. In general, llGetSubString or llList2List return the whole thing when the starting index is grater than the last element in the list.
There were a number of spots with that hazard, so fix all we found.
It should help ensuring that you have the right script contents when saving. It also provides a short description, to get a rough idea of what it's about.
The only code change is the product name in [AV]xcite! to add a TM symbol to Xcite!
Use a global variable instead. In SL, the optimizer does the replacement automatically, therefore no memory penalty is incurred. In OpenSim, the variable is assigned at run time with llUnescapeURL. The OpenSim separator is also changed from U+001F to U+007F, because most control characters under U+0020 cause problems with exporting when present in script memory.
Fixes#41.
We knew that in OpenSim, the bitwise NOT "~" operator's precedence was broken (see <http://opensimulator.org/mantis/view.php?id=3268>), but it appears that the same bug affects the logical NOT operator "!" as well, so it needs parentheses when followed by more expressions.
In OpenSim, llGetLinkNumber() returns 0 for single prims with a sitting avatar. On top of that, in older versions llGetLinkKey returns NULL_KEY for link number 0, and in newer ones it does so for link number 1.
The present changes try to make sure that under either situation, the code behaves correctly for the core scripts.
- Add Makefile and release creation instructions.
- Add a simple Python program to automate the OpenSim conversion based on markings in the code.
- Add .gitignore entries for the generated files.
- Add parentheses around assignments as required by OpenSim. This is done only to the core scripts.
- OpenSim isn't compatible with SL when there are conditions of type key. Those are all converted. Conditions of other types, except integer, are expanded for clarity and optimization, as they generate the same or better code that way, and currently the optimizer can do a better job when they are expanded.
- Floats in scientific notation need a dot.
- llParseStringXXXX doesn't work the same in OpenSim as in SL, when the separator is an Unicode codepoint that doesn't represent a character. For that reason, the internal separator, which is U+FFFD ("Replacement Character") is changed automatically by the Python program to U+001F (Unit Separator control character). For further safety, function strReplace is altered to use osReplaceString instead of llParseStringKeepNulls/llDumpList2String.
Furthermore, the ~ operator has the wrong precedence in OpenSim, but that was handled by a previous commit. Note that appearances of the ~ operator that were not preceded by a ! have only been replaced in the core scripts.
Many of them cause problems with OpenSim.
All of them are applied by the optimizer, except (str="")+str, which produces gains only in LSO, not in Mono, and doesn't work in OpenSim.
A few are undone only for clarity, relying on the fact that the optimizer will apply them again.
- Add spaces between functions.
- Add comments on numeric codes and a few others.
- Reformat lists and one function.
- Remove spaces at EOL.
- Use actual copyright symbol.
