>I saw it in the header file, but actually testing it on Solaris 10 Sparc >showed that the address was different as well in different threads. Can >you try on you system as well?
Actually, my system is OpenIndiana, but you're right. The address of ___errno() function is same, but the return value of ___errno() (that is, the pointer to thread-local errno) is different. Sorry for the confusion. Tsuyoshi SASAMOTO nazon...@miobox.jp