llDumpList2String has changed its behaviour with respect to minus zero. Now it converts -0.0 to a string without the minus sign.
While testing this, we noticed several mismatches in the float to string conversions; the existing routine did not properly convert some values because as we discovered later, it is subject to double rounding; one of them is the built-in round-to-nearest-or-even while getting the first 7 significant digits, and the other is just an increment when the digit is a 5 or more, so round to nearest, ties away from zero, and is performed on the digit past the five or six visible digits that LSL shows.
The new code is a tad easier to understand and more robust.
A first variant of the new code is left commented out for history's sake, and will be removed in the next commit.
This annoyance and discrepancy with LSO was finally fixed by Linden Lab. The change has prompted some modifications to the test suite to accommodate for the new results. A further improvement has been to make these tests more friendly to be run in SL, making it easier to verify the results.
Fixes#17.
Reported by SaladDais@users.noreply.github.com - thanks!
They are transformed by the scanner to the identifier `inline`, which is how the parser identifies it. This solves the comment problem, but it results in a funny side effect. Now, in inline mode, /*pragma inline*/ will always be the identifier `inline`, therefore this is valid:
integer /*pragma inline*/ = 5;
llOwnerSay((string)inline); // will say 5
Not overly elegant, but better than making up a specific token or declaring comments as tokens or the like.
1. When the last statement of a function is a RETURN statement which is syntactically required, it could still be deleted.
2. The child of a RETURN statement could be removed if the statement was not executed.
This commit fixes both issues.
Bug report and test case provided by @Tonaie. Fixes#14.
E.g. llSubStringIndex(...) > -1 was converted into !~llSubStringIndex(...) which is incorrect. We even had a test case for it... with a wrong expected response file.
Bug report and test case by Sinha Hynes (thanks!)
vector * quaternion: Simplified by precalculating the repeated products and removing math.fsum.
quaternion * quaternion: Add more F32's to get a result closer to SL's (still not there, but we're definitively closer now).
We left out quaternion/quaternion in the tests, and the q*q test was not general enough (had many zeros). Remedied that.
The algorithm for adding parentheses around unary operators was not working properly. It converted a * (-b) * c into a * -b * c, which LSL handles as a * -(b * c).
Fix and add test cases for that. One of the test cases shows an example where the difference matters: 0 * (-1e20) * 1e20 should result in 0.0, but if wrongly parenthesized, it gives NaN, because 1e20*1e20 gives infinity due to float overflow, and minus infinity times 0 is indeterminate.
The addition of parentheses has been improved, but it still does not eliminate every redundant parenthesis.
Also fix the horrendous typo of using "operands" where it should be "operators".
In the same places as state changes are allowed, i.e. in places where a parent of the AST node is a WHILE/DO/FOR or an IF without ELSE, it's allowed to use return statements with expressions that return void, e.g. llDie(), provided the function itself is declared as returning void.
The construction, when found, is rewritten to '{<void expression>; return;}' because the optimizer is not designed to deal with these monsters.
We've renamed the variable SuspiciousStSw to PruneBug, because it's used for both purposes now, though a better name might have been PruneBugPendingChecks, because these are only errors if the IF has an ELSE. We've also added the exception to raise as part of the data stored in the list.
Per report by Tonaie Resident.
Revert "Add support for C11-style _Pragma operator on processpre".
Revert "Add unit test for the _Pragma operator".
This reverts commits 31fcb331c7 and e261ac2121.
This should rather be the job of the preprocessor, which should generate #pragma lines. gcc does that.
This is a first try at redundant jump removal (jumps that target the very next instruction). It's too basic in several ways.
- The statement is replaced by a ';' instead of removed.
- If the jump was the only statement in an if, when the if becomes empty, it's not folded.
- Jumps that are last in the 'then' branch of if+else are not visible. This would need either to track multiple last statements, or to have some means to anticipate what the next statement is at every statement. A Control Flow Graph would help a lot.
- When a label is immediately followed by a jump, all jumps to that label should target the destination of that jump if it's in scope. Added to TODO.
- It misses some optimizations when not expanding WHILE and FOR into IF/JUMP.
Moving everything to an earlier stage would help with some of these, especially with ';' and 'if' folding. Unconditionally expanding WHILE and FOR would also help.
This test suite has been in use for a long time now, in place of the obsolete and unmanageable testparser.py and testfuncs.py. It verifies the complete optimizer output to stdout and stderr, to ensure that the output matches the expectations.
See unit_tests/README.txt for more info.
