diff options
| author | Joseph Myers <joseph@codesourcery.com> | 2015-11-24 22:21:59 +0000 |
|---|---|---|
| committer | Joseph Myers <joseph@codesourcery.com> | 2015-11-24 22:21:59 +0000 |
| commit | a7f0c5ae4184916f0e145de3aefc794bf2e280ad (patch) | |
| tree | d8b6c7ed00505c051a0a67b0b8f3810b6c385b03 /wcsmbs/Makefile | |
| parent | 869c2f1eae93e789fb2deb4c82f3cb2ee2aac9c5 (diff) | |
| download | glibc-a7f0c5ae4184916f0e145de3aefc794bf2e280ad.tar.gz glibc-a7f0c5ae4184916f0e145de3aefc794bf2e280ad.tar.xz glibc-a7f0c5ae4184916f0e145de3aefc794bf2e280ad.zip | |
Fix strtod ("NAN(I)") in Turkish locales (bug 19266).
The implementations of strtod and related functions use locale-specific conversions to lower case when parsing the contents of a string NAN(n-char-sequence_opt). This has the consequence that NAN(I) is not treated as being of that form (only the initial NAN part is accepted). The syntax of n-char-sequence directly maps to the ASCII letters, digits and underscore as in identifiers, so it is unambiguous that all ASCII letters must be accepted in all locales. This patch, relative to a tree with <https://sourceware.org/ml/libc-alpha/2015-11/msg00258.html> (pending review) applied and depending on that patch, fixes this problem by checking directly for ASCII letters. This will have the side effect of no longer accepting 'İ' (dotted 'I') inside NAN() in Turkish locales, which seems appropriate (that letter wouldn't have been interpreted as having any meaning in the NaN payload anyway, as not acceptable to strtoull). Tested for x86_64 and x86. [BZ #19266] * stdlib/strtod_l.c (____STRTOF_INTERNAL): Check directly for upper case and lower case letters inside NAN(), not using TOLOWER. * stdlib/tst-strtod-nan-locale-main.c: New file. * stdlib/tst-strtod-nan-locale.c: Likewise. * stdlib/Makefile (tests): Add tst-strtod-nan-locale. [$(run-built-tests) = yes] ($(objpfx)tst-strtod-nan-locale.out): Depend on $(gen-locales). ($(objpfx)tst-strtod-nan-locale): Depend on $(libm). * wcsmbs/tst-wcstod-nan-locale.c: New file. * wcsmbs/Makefile (tests): Add tst-wcstod-nan-locale. [$(run-built-tests) = yes] ($(objpfx)tst-wcstod-nan-locale.out): Depend on $(gen-locales). ($(objpfx)tst-wcstod-nan-locale): Depend on $(libm).
Diffstat (limited to 'wcsmbs/Makefile')
| -rw-r--r-- | wcsmbs/Makefile | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/wcsmbs/Makefile b/wcsmbs/Makefile index c1bb9375b7..773cfdb932 100644 --- a/wcsmbs/Makefile +++ b/wcsmbs/Makefile @@ -47,7 +47,7 @@ strop-tests := wcscmp wcsncmp wmemcmp wcslen wcschr wcsrchr wcscpy wcsnlen \ wcscspn wmemchr wmemset tests := tst-wcstof wcsmbs-tst1 tst-wcsnlen tst-btowc tst-mbrtowc \ tst-wcrtomb tst-wcpncpy tst-mbsrtowcs tst-wchar-h tst-mbrtowc2 \ - tst-c16c32-1 wcsatcliff tst-wcstol-locale \ + tst-c16c32-1 wcsatcliff tst-wcstol-locale tst-wcstod-nan-locale \ $(addprefix test-,$(strop-tests)) include ../Rules @@ -64,6 +64,7 @@ $(objpfx)tst-mbrtowc2.out: $(gen-locales) $(objpfx)tst-wcrtomb.out: $(gen-locales) $(objpfx)wcsmbs-tst1.out: $(gen-locales) $(objpfx)tst-wcstol-locale.out: $(gen-locales) +$(objpfx)tst-wcstod-nan-locale.out: $(gen-locales) endif CFLAGS-wcwidth.c = -I../wctype @@ -95,3 +96,5 @@ CPPFLAGS += $(libio-mtsafe) # We need to find the default version of strtold_l in stdlib. CPPFLAGS-wcstold_l.c = -I../stdlib + +$(objpfx)tst-wcstod-nan-locale: $(libm) |