Sergei Gorelkin wrote:
Sergei Gorelkin wrote:
I'll be digging further in.
Here goes. I modified the test so that it fails reliably, and patched
the compiler so that is doesn't fail again.
It was a bit tricky to get a string with refcount=1 without telling the
compiler to take its address.
As for the patch, I could not invent anything more clever than comparing
the assignment target with all function parameters one by one. Also
typeconversion nodes may have to be handled.
This patch is not good enough. Like you already say typeconversions are
not handled. And what about subscriptions or dereferences. Minor remark
is the name of the method which is too generic and not making explicit
what it does.
_______________________________________________
fpc-devel maillist - fpc-devel@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-devel
