[Webkit-unassigned] [Bug 15416] dtoa Falls into infinite loop on ARMEL

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=15416





------- Comment #18 from martinwguy at yahoo.it  2008-09-03 01:46 PDT -------
Hi!
   I found this bug report while searching for something else. I had to wrangle
ARM FP variants over the last few years for wiki.debian.org/ArmEabiPort so can
shed some light...
   Unfortunately there is no single define to isolate the bizarre
45670123-order doubles as used in ancient ARM FPA and the old-abi FP emulator,
but to the code in Platform.h:

#elif !defined(__ARM_EABI__) && !defined(__ARMEB__) && !defined(__VFP_FP__)
#define WTF_PLATFORM_MIDDLE_ENDIAN 1

you might add && !defined(__MAVERICK__), which is the other true-little-endian
FP coprocessor from Cirrus. __VFP_FP__ is also defined if soft-float is in use,
so that case is alrady covered correctly by this code.

Also, the following phrase
#if !defined(__ARM_EABI__)
#define WTF_PLATFORM_FORCE_PACK 1
#endif
is suspect, since ARMs require data accesses to be aligned to a multiple of
their size for 2- and 4-byte words regardless of ABI; the only difference is
that in EABI, doubles should be 8-byte-aligned instead of 4.
However WTF_PLATFORM_FORCE_PACK is not used anywhere in the rest of WebKit, so
it looks redundant now.

Cheers!


-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.