|
|
|
@ -456,3 +456,37 @@ bug fixes: |
|
|
|
- strtoull wrongly rejecting the highest 16 possible values as overflow |
|
|
|
- math.h constant expression fixes for INFINITY/NAN/etc. |
|
|
|
- scanf mishandling of "0" with "%x" |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0.8.8 release notes |
|
|
|
|
|
|
|
new feature: |
|
|
|
- major math correctness and performance improvements |
|
|
|
- many math functions implemented in asm for i386 |
|
|
|
- some math functions (mostly long double) in asm for x86_64 |
|
|
|
- new floating point parser/converter with correct rounding |
|
|
|
- implement wcstod, wcstof, and wcstold |
|
|
|
- new scanf implementation - cleaner, faster, more correct |
|
|
|
- minimal/incomplete strfmon implementation |
|
|
|
|
|
|
|
compatibility: |
|
|
|
- header fixes for c++ |
|
|
|
- regex code resync with TRE; support common regex extensions |
|
|
|
- support for compiling apps with gcc's -funsigned-char |
|
|
|
- sysconf now returns dynamic limits for open files, processes |
|
|
|
- give dlerror proper error status stickiness |
|
|
|
- make alloca work even with -fno-builtin |
|
|
|
|
|
|
|
critical security fixes: |
|
|
|
- stack-based buffer overflow in fprintf on unbuffered files |
|
|
|
|
|
|
|
other bug fixes: |
|
|
|
- rare gcc register allocation (miscompilation) bug in syscall wrappers |
|
|
|
- printf was rejecting the valid (but redundant) %lf format specifier |
|
|
|
- fixed big data bloat (missing const) in math functions |
|
|
|
- many math fixes related to floating point exceptions and rounding |
|
|
|
- corrected DECIMAL_DIG definitions |
|
|
|
- tgammal was wrongly setting global signgam |
|
|
|
- crash in wordfree with uninitialized we_offs |
|
|
|
- fix wordexp not null-initializing the we_offs initial slots |
|
|
|
|
Rich Felker
14 years ago