This is the mail archive of the newlib@sourceware.org mailing list for the newlib project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
Hi, A recent change to g++ means that we now put out calls to __sync_synchronize for C++ guard variables for versions of the architecture that do not have a barrier instruction. Instead of ignoring the problem in the compiler and working around it by providing a stub in libgcc, provide a weak definition here in order for testing to succeed but to highlight this for users provide a link time warning when this is used. Additionally libgfortran now pulls in libbacktrace which expects to be able to use atomic accesses where required. This creates problems when testing in bare-metal configurations. This means though that developers and testers for arm-none-eabi will have to add something to prune additional output in their site.exp's with an additional implementation of prune_warnings to prune the warnings but this is more prudent than just putting out silent calls to __sync_synchronize. Tested this for arm-none-eabi with --with-multilib-list=aprofile with suitable multilibs. Thoughts ? Ok ? Ramana <DATE> Ramana Radhakrishnan <ramana.radhakrishnan@arm.com> * libc/machine/arm/sync_synchronize.c: New file. * libc/machine/arm/Makefile.am: Add sync_synchronize.c * libc/machine/arm/Makefile.in: Regenerate.
Attachment:
adjust-newlib.txt
Description: Text document
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |