On 09/08/2016 01:03 PM, Martin Sebor wrote:
Attached is another update to the patch to address the last round
of comments and suggestions, most notably to:
* implement the checking of the implementation limit of 4,095 on
the output of a single directive to allow for the Glibc failure
due to ENOMEM (the patch issues a warning and disables the
optimization when this happens)
* implement checking for exceeding INT_MAX bytes (warn and disable
optimization)
* call set_range_info when the return value optimization is not
possible
* remove code to work around tree-optimization/71831 (now on
trunk)
The -fprintf-return value optimization is still disabled. GCC
successfully bootstraps with it and most tests pass but there's
a failure in the Fortran libgomp tests that I am yet to figure
out.
I'm just now getting back to this (sorry for the long delay). I see
there's another version after this one. Did the version from 9/11/2016
address the Fortran libgomp test issue?
jeff
