| commit | 59f2c1749f77c1ef30aa0db3802a8b77df1037fd | |
| author | Mark Williams <mwilliams@fb.com> | |
| Thu, 16 May 2013 19:46:47 +0000 (16 12:46 -0700) | ||
| committer | Sara Golemon <sgolemon@fb.com> | |
| Mon, 20 May 2013 20:52:29 +0000 (20 13:52 -0700) | ||
| tree | 0c623b6098549baedae878f0ce101299b0d5fd42 | treesnapshot (tar.gz zip) |
| parent | ca4d38d5ff82d9d83fd5fa5dd92d125c1e6c48d2 | commitdiff |
Fewer varEnvs in generators
If the original function has a variable that the generator does not,
we create a varenv. Its not clear that that should ever happen, but
it *was* happening because:
- closures get a magic variable called 0Closure, but generators
from closures did not.
- unused parameters didnt get put into the variable table, so
generators with unused parameters got a varenv
If the original function has a variable that the generator does not,
we create a varenv. Its not clear that that should ever happen, but
it *was* happening because:
- closures get a magic variable called 0Closure, but generators
from closures did not.
- unused parameters didnt get put into the variable table, so
generators with unused parameters got a varenv
| hphp/compiler/analysis/emitter.cpp | diffblobblamehistory | |
| hphp/compiler/statement/method_statement.cpp | diffblobblamehistory |