Chromatic <[EMAIL PROTECTED]> wrote: > On Thu, 2004-04-15 at 09:18, Leopold Toetsch wrote:
>> Ah. Relikt of Jeff's patch. If that constant got reused elsewhere, e.g. >> as a method name, it were one too short. > Confirmed. Thanks, Leo! Good. > Would a test patch such as the following be good to catch regressions, I didn't boil it down to a test. It was just a look at the patch, that *could* have caused the bug. Simple usage of single quotes was ok. Multiple usage too (constant folding jumps in). But name mangling + shortened original symbol could cause the problem. Just replacing double quotes with single is and was working. Only the combination of different usage of one string constant could have triggered the bug. For a test you need (probably) usage of that string constant as a single quoted string and as a method name - maybe in a different namespace. leo