This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: [c++0x] <date_time> implementation, take 4
Hi Pedro,
I thought this test failed because of new symbols added.
The test is specifically crafted to allow for symbol *additions*, the
best type of compatibility we can deliver between releases while not
blocking development completely. That said, a tricky situation happens
when you **inadvertently** export additional symbols, that this you
export those symbols at an old revver, not the most recent one.
abi_check lately tests for that annoying situation too (and that's why
was recently failing: for some reason, inlining of two small functions
was not happening anymore and two symbols, inadvertently matching old
patterns in the linker script, were inadvertently exported)
My working copy is not up to date: Last Changed Rev: 133070
Humm, I'm worried because Honza fixed that issue at Rev. 132854...
I'am afraid you are stumbling in the tricky situation above: some new
symbols are inadvertently matching old patterns in the linker script...
Please try to make progress, inspect the libstdc++.log file, abi_check
outputs all the required diagnostic information.
Paolo.