Re: [stdc++] Mark global new/delete operators as weak symbols

2013-09-04 Thread Mike Stump
On Sep 4, 2013, at 7:30 AM, Romain Geissler wrote: >> 2013-08-20 Romain Geissler >> >>* config/os/gnu-linux/os_defines.h: (_GLIBCXX_WEAK_DEFINITION): Define. Strikes me as wrong. Using weak should be autoconfed or driven by the compiler and then respected by the library. With that chan

Re: [stdc++] Mark global new/delete operators as weak symbols

2013-09-04 Thread Romain Geissler
Ping 2013/8/20 Romain Geissler : > Hi, > > According to the C++ standard §17.4.3.4(2,3) the global new and delete > operator implementation can be replaced by the program. For example, > tcmalloc does redefine them. On ELF targets, it is currently OK if you > link libstc++ dynamically. However, if

[stdc++] Mark global new/delete operators as weak symbols

2013-08-19 Thread Romain Geissler
Hi, According to the C++ standard §17.4.3.4(2,3) the global new and delete operator implementation can be replaced by the program. For example, tcmalloc does redefine them. On ELF targets, it is currently OK if you link libstc++ dynamically. However, if you try to link both tcmalloc and stdc++ sta