On 13/04/15 20:37 +0200, Jakub Jelinek wrote:
On Mon, Apr 13, 2015 at 07:28:30PM +0100, Jonathan Wakely wrote:
This fixes a regression I introduced by not exporting some of the
fstream constructors (the C2 base object ones) when using the new
std::string. Sorry for not finding this myself before the RC.
All the recently-updated baseline_symbols.txt files need updating. I
did this with sed and sort, and verified that the x86_64-linux-gnu and
powerpc64-linux-gnu ones are identical (except for the TLS call_once
symbols) to the output of 'make new-abi-baseline'. I have not checked
the baselines for other arches, only inspected the diff visually.
This really needs to go on trunk and gcc-5-branch.
This is ok for 5.1, but please commit it to gcc-5-branch soon, so that
people have time to retest on their architectures.
Committed to trunk and gcc-5-branch.
