From c1e2141f7500a6d9282fae382c6ddd6d0f5d59cb Mon Sep 17 00:00:00 2001 From: Xavier Leroy Date: Tue, 31 Jan 2017 13:46:14 +0100 Subject: Update Changes for release 3.0 --- Changelog | 41 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 39 insertions(+), 2 deletions(-) (limited to 'Changelog') diff --git a/Changelog b/Changelog index 59f10c2f..9332ac43 100644 --- a/Changelog +++ b/Changelog @@ -1,5 +1,5 @@ -Working version -=============== +Release 3.0, 2017-02-xx +======================= Major improvements: @@ -18,6 +18,43 @@ Major improvements: above.) Support for x86-64 is currently available for Linux and MacOS X. (Run the configure script with 'x86_64-linux' or 'x86_64-macosx'.) +Language features: + +- Support for anonymous structures and unions as members of + structures or unions. (ISO C11, section 6.7.2.1, para 13 and 19.) +- New built-in functions for ARM and PowerPC: + __builtin_ctz, __builtin_ctzl, __builtin_ctzll + (count trailing zeros, 32 and 64 bits). + +Usability: + +- Added options -Wxxx and -Wno-xxx (for various values of "xxx") + to control which warnings are emitted. +- Support response files where additional command-line arguments can + be passed (syntax: @file). +- Improved wording of warning and error messages. + +Code generation: + +- Support for ARM target processors in big-endian mode. +- Optimize 64-bit integer division by constants. + +Bug fixing: + +- Issue #155: on ARM, assembly errors caused by large jump tables for + "switch" statements and overflow in accessing constant pools. +- Issue #151: large inductive definition causes a fatal error in + 32-bit versions of Coq. +- Issue #143: handle "%lf" printf() format in the reference interpreter +- Issue #138: struct declarations in K&R function parameters were ignored. +- Issues #110, #111, #113, #114, #115, #119, #120, #121, #122, #123, #124, + #125, #126, #127, #128, #129, #130, #133, #138, #144: various cases + of internal errors and failed assertions that should have been + proper errors instead. +- For __builtin_memcpy_aligned, size and alignment arguments of 64-bit + integer type were causing a fatal error on a 32-bit target. +- ARM and x86 ports: wrong register allocation for some calls to + function pointers. Release 2.7.1, 2016-07-18 -- cgit