Re: strtod bugs

[Bruno Haible]

Eric Blake wrote:
> test-strtod.c is all I had; but I intentionally made it catch all errors
> rather than aborting on the first

Ah, I see. Sorry, I had not seen this detail.

Find attached the results on a dozen of platforms.

I integrated these test results into test-strtod.c as comments (since we don't
yet have a matrix like in m4/printf.m4), using a script like this:

sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* glibc-2.3.6 
*/|;},' < test-strtod-glibc-2.3.6 > sed1
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* MacOS X 10.3 
*/|;},' < test-strtod-macosx-10.3.9 > sed2
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* FreeBSD 6.2 
*/|;},' < test-strtod-freebsd-6.2 > sed3
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* NetBSD 3.0 
*/|;},' < test-strtod-netbsd-3.0 > sed4
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* OpenBSD 4.0 
*/|;},' < test-strtod-openbsd-4.0 > sed5
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* AIX 4.3 */|;},' < 
test-strtod-aix-4.3.2 > sed6
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* AIX 5.1 */|;},' < 
test-strtod-aix-5.1 > sed7
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* HP-UX 11.00 
*/|;},' < test-strtod-hpux-11.00 > sed8
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* HP-UX 11.11 
*/|;},' < test-strtod-hpux-11.11 > sed9
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* IRIX 6.5 */|;},' 
< test-strtod-irix-6.5 > sed10
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* OSF/1 4.0 */|;},' 
< test-strtod-osf1-4.0d > sed11
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* OSF/1 5.1 */|;},' 
< test-strtod-osf1-5.1 > sed12
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* Solaris 2.4 
*/|;},' < test-strtod-solaris-2.4 > sed13
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* Solaris 2.5.1 
*/|;},' < test-strtod-solaris-2.5.1 > sed14
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* Solaris 9 */|;},' 
< test-strtod-solaris-2.9 > sed15
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* Solaris 10 
*/|;},' < test-strtod-solaris-2.10 > sed16
sed -e 's,test-strtod.c:,,' -e 's,: assertion failed,{s|$| /* mingw */|;},' < 
test-strtod-mingw > sed17
cat gnulib-git/tests/test-strtod.c \
  | sed -f sed1 | sed -f sed2 | sed -f sed3 | sed -f sed4 | sed -f sed5 \
  | sed -f sed6 | sed -f sed7 | sed -f sed8 | sed -f sed9 | sed -f sed10 \
  | sed -f sed11 | sed -f sed12 | sed -f sed13 | sed -f sed14 | sed -f sed15 \
  | sed -f sed16 | sed -f sed17 > gnulib-git/tests/test-strtod.c.new

Then I also updated the doc accordingly. Notable disagreements:
  - I could not see an indication of the "returns a wrong end pointer on some
    platforms: Solaris 2.4" problem, so I removed this;
  - You seem to have seen problems on Solaris 8 that I did not see. What can
    be the reason? Did you use gnulib's signbit and isnan functions? I used
    the ones from the system, or trivial platform-specific replacements.
  - Changed wording: 0.0 is not considered a "positive value". The usual
    definition of "positive" is "x > 0", and "negative" is "x < 0". Yes, a
    "negative zero" is not negative; that's actually a misnomer.


2008-03-30  Bruno Haible  <address@hidden>

        * tests/test-strtod.c: Add comments about which assertion fails on which
        platform.
        * doc/posix-functions/strtod.texi: Add info about many more platforms.

--- doc/posix-functions/strtod.texi.orig        2008-03-31 01:56:47.000000000 
+0200
+++ doc/posix-functions/strtod.texi     2008-03-31 01:56:40.000000000 +0200
@@ -16,53 +16,52 @@
 Old versions of Linux.
 
 @item
-This function returns a wrong end pointer on some platforms:
-Solaris 2.4.
+This function returns a wrong end pointer on some old platforms.
 
 @item
 This function fails to do a valid parse of @samp{-0x} on some
 platforms:
-glibc 2.4, Cygwin < 1.5.25-11.
+glibc 2.4, MacOS X 10.3, FreeBSD 6.2, Cygwin < 1.5.25-11.
 
 @item
 This function fails to parse Infinities and plain NaNs on some platforms:
-Solaris 8, Mingw, OpenBSD 4.0.
+OpenBSD 4.0, HP-UX 11.11, IRIX 6.5, OSF/1 5.1, Solaris 9, mingw.
 
 @item
 This function fails to parse NaN() on some platforms:
-Solaris 8, Mingw, OpenBSD 4.0, Cygwin < 1.5.25-11.
+glibc-2.3.6, MacOS X 10.3, FreeBSD 6.2, OpenBSD 4.0, AIX 5.1, HP-UX 11.11, 
IRIX 6.5, OSF/1 5.1, Solaris?, Cygwin < 1.5.25-11, mingw.
 
 @item
 This function fails to parse NaN(n-char-sequence) on some platforms:
-Solaris 8, Mingw, OpenBSD 4.0.
+OpenBSD 4.0, HP-UX 11.11, IRIX 6.5, OSF/1 5.1, Solaris?, mingw.
 
 @item
 This function returns the wrong end pointer when parsing
 NaN(n-char-sequence) on some platforms:
-glibc 2.4.
+glibc-2.4, OpenBSD 4.0, AIX 5.1, HP-UX 11.11, IRIX 6.5, OSF/1 5.1, Solaris?, 
mingw.
 
 @item
 This function fails to parse C99 hexadecimal floating point on some
 platforms:
-Solaris 8, Mingw, OpenBSD 4.0.
+NetBSD 3.0, OpenBSD 4.0, AIX 5.1, HP-UX 11.11, IRIX 6.5, OSF/1 5.1, Solaris 
10, mingw.
 @end itemize
 
 Portability problems not fixed by Gnulib:
 @itemize
 @item
-This function returns a positive value for negative underflow on some
+This function returns +0.0 (not -0.0) for negative underflow on some
 platforms:
-glibc 2.7, Mingw, Cygwin.
+glibc 2.7, Cygwin, mingw.
 
 @item
 This function cannot distinguish between ``nan'' and ``-nan'' on some
 platforms:
-glibc 2.7.
+glibc 2.7, IRIX 6.5, OSF/1 5.1, mingw.
 
 @item
 This function fails to correctly parse very long strings on some
 platforms:
-Mingw, Cygwin.
+MacOS X 10.3, FreeBSD 6.2, NetBSD 3.0, OpenBSD 4.0, IRIX 6.5, OSF/1 5.1, 
Cygwin, mingw.
 
 @item
 The replacement function does not always return correctly rounded results.

test-strtod.tar.gz
Description: application/tgz