> What I am really hearing in this thread is some people have interest in a > way to easily access the __COMPILER_HALT_OFFSET__
Well, not exactly for me : I'm not interested strictly on getting the offset. I'm more interested in doing generic static code analysis, and part of that is getting the halt offset, amongst many other data coming from tokens. The big pb with the fix applied since 5.3.6 is that we can not do that anymore without many complications. So in this way, I agree with Nikita, the second aspect of the patch (getting the binary data in a T_INLINE_HTML) is not a necessity, thought that may help. But the first aspect of the patch (counting 3 semantic tokens after T_HALT_COMPILER) is really needed. That would just give us back the power/feature we have lost since 5.3.6. So my request would be : pleeeease apply the patch counting 3 semantic tokens for 5.4 asap, then we can discuss the remaining later for the shake of completeness of the discussion. Nicolas