The current test code for complex numbers is just Plain Bad (TM).
Someone else than me :-) should rewrite the test to use more
modern testing frameworks. Note: large parts of the current
test are data-driven, I have no experience on how well e.g.
Test::More suits that, if it doesn't (I am
The attached patch (#1) brings several pending updates to Math::Complex
and Math::Trig.
- Complex: fix for the [perl #31117]: atan2(0, i) now works,
as do all the (computable) complex argument cases (I adopted
the Mathematica definition)
- Complex: fixes for
Rafael Garcia-Suarez wrote:
Jarkko Hietaniemi wrote:
Both applied to blead as #25414, thanks. But may I suggest that next time,
you run tests before sending them ? Just look again at your next_test()
Ooops. In my defense I can say I did
perl -wIlib lib/Math/Complex.t|grep not
(and came
There is a small technical bug in sub great_circle_destination {...}.
An additional (optional) argument, RHO, must be passed to the
subroutine. Otherwise, the calculation will always be based on the
I am confused. On a sphere the destination point d (radians)
one ends up after going from
Hence, I'd suggest patching the behaviour one way or another. There's
three things I'd consider sensible to do:
- atan2($z1, $z2) = atan($z1/$z2) (same order as with the normal atan2)
- remove the atan2($z1, $z2) alltogether and just carp() instead.
- atan2($z1, $z2) =
The external gzip support in IO::Zlib is all Jarkko's work and
nothing to do with me I'm afraid.
OK, CC'ed Jarkko as well :-)
Jarkko, you been paying attention?
Not awfully so, I don't read p5p actively. [rereading the thread]
Off-hand I'd say the easy thing to do would be just skip
This bit:
: This is not correct. Obviously, 0/i is the same as 0/1
: which is 0. Thus atan2(0,i) == atan2(0,1) == atan(0) == 0
seems to be incorrect... if one believes the Mathematica definition
atan2(0, i) is pi/2, not zero.
Ahem, sorry... it seems that Mathematica's ArcTan is
While I certainly agree that a runtime divide-by-zero is not the right
thing to do, I cannot quite bring myself to agree on the proposed fix.
In my books atan2() is not well-defined for complex arguments at all,
and even if in some special cases it might have a reasonable definition,
I still
Jarkko Hietaniemi wrote:
While I certainly agree that a runtime divide-by-zero is not the right
thing to do, I cannot quite bring myself to agree on the proposed fix.
In my books atan2() is not well-defined for complex arguments at all,
and even if in some special cases it might have
H.Merijn Brand wrote:
On 15 Aug 2005 21:05:22 -0700, Gisle Aas [EMAIL PROTECTED] wrote:
I noticed that our Solaris builds ends up with a Time::HiRes module
that does not support nanosleep. This happens because the Solaris
hints file for Time::HiRes tries to use the POSIX module that is not
The attached patch removes some duplicated duplicated words.
Note that some changes affect the perlintern.pod while some
affect pods of external origin like Encode and the FAQ; many
are applicable to maint, too.
--- ext/Encode/lib/Encode/Supported.pod.dist2005-08-07 12:52:14.0
H.Merijn Brand wrote:
On Sun, 07 Aug 2005 13:28:19 +0300, Jarkko Hietaniemi [EMAIL PROTECTED]
wrote:
The attached patch removes some duplicated duplicated words.
Note that some changes affect the perlintern.pod while some
affect pods of external origin like Encode and the FAQ; many
implement a pragma for shifting
of scalars as bitvecs.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
Jarkko Hietaniemi wrote:
A better way to declare strict ANSI-C-ness in Tru64 is to use
the -c99 for ccflags (but still -std1 for the lddlflags), but
of course use the -c99 only if it is available.
A useful feature that becomes available with -c99 is the __func__,
which is handy
A better way to declare strict ANSI-C-ness in Tru64 is to use
the -c99 for ccflags (but still -std1 for the lddlflags), but
of course use the -c99 only if it is available.
A useful feature that becomes available with -c99 is the __func__,
which is handy for the -DPERL_MEM_LOG.
tru64h.pat.gz
Steve Peters wrote:
On Wed, Jul 20, 2005 at 02:40:54PM +0300, Jarkko Hietaniemi wrote:
My (?:m(?:is)?)?adventures in the world of POSIX signals continue.
Now to followup on my own patch of two years ago [1], here is support
for SA_SIGINFO (see man sigaction), which allows additional
all these years, I think I am in a minority,
and I can live with it. *IF* someone wants to fix this, maybe a pragma.
Or maybe borgify Bit::Vector :-)
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack
I agree. Mac OS X tells me nothing apart from what's in signal.h,
man siginfo in Solaris and man sigaction in AIX and Tru64 are quite
helpful. But I think for the time being the attached patch will have
to do: only signo, code, errno are decoded (seems the be the smallest
supported set from
As suggested by Andy Dougherty.
NSIG.pat.gz
Description: GNU Zip compressed data
My (?:m(?:is)?)?adventures in the world of POSIX signals continue.
Now to followup on my own patch of two years ago [1], here is support
for SA_SIGINFO (see man sigaction), which allows additional information
to be passed on to the signal handlers (beyond the signal number).
I am not entirely
Joshua Hoblitt wrote:
On Sun, Jun 15, 2003 at 09:14:16PM +0300, Jarkko Hietaniemi wrote:
This just isn't cricket for non-GCC compilers (Solaris, AIX, and Tru64
claim a syntax error, IRIX seems to tolerate it). That a function
call (Perl_doing_taint in this case) gets expanded to func(a,b
The attached patch introduces %POSIX::SIGRT which gives access
to the POSIX real time signals SIGRTMIN...SIGRTMAX, with the
right POSIXly moves. It also plugs a hole the size of a coredump
I accidentally run into while fooling around with the patch, try
perl -MPOSIX -e 'sigaction(123,0)'
Jarkko Hietaniemi wrote:
The attached patch introduces %POSIX::SIGRT which gives access
to the POSIX real time signals SIGRTMIN...SIGRTMAX, with the
right POSIXly moves. It also plugs a hole the size of a coredump
I accidentally run into while fooling around with the patch, try
perl
The attached small patch is poor man's valgrind: it injects a dose
of logging into Newx() et alia so that if PERL_MEM_LOG is defined in
compile-time, all Newx() et alia calls go through logging functions,
which most importantly pass in the source code file and linenumber.
By default the logging
Dave Mitchell wrote:
On Fri, Jul 08, 2005 at 09:45:32AM +0300, Jarkko Hietaniemi wrote:
I won't have time to look into this anytime soon, but I think the fix
for the second case shouldn't be too hard to find. First of all, if
either the matcher or the matchee are UTF-8, we should never ever
Dave Mitchell wrote:
On Sun, Jul 10, 2005 at 01:03:10PM +0300, Jarkko Hietaniemi wrote:
The attached small patch is poor man's valgrind: it injects a dose
of logging into Newx() et alia so that if PERL_MEM_LOG is defined in
compile-time, all Newx() et alia calls go through logging functions
it turns out perl is totally borked for
That bugs of this level are found only now is telling something about
how much people actually use the Unicode support of Perl. Well, and
something about how suckily I tested it.
The patch below fixes the first case: the fault was in ibcmp_utf8(),
Dave Mitchell wrote:
On Mon, Jul 04, 2005 at 12:19:28AM +0300, Jarkko Hietaniemi wrote:
Just rotate the bits 90 degrees clockwise, shake well, and look against
a strong light. Oh, right, there are no bits to rotate. Try these bits.
Under a full moon while standing with one trouser leg
Jarkko Hietaniemi wrote:
cc: Warning: perl.h, line 4175: In the initializer for
PL_vtbl_arylen.svt_get, the referenced type of the pointer value
Perl_magic_getarylen is function (pointer to struct sv, pointer to
const struct magic) returning int, which is not compatible with
function (pointer
Jarkko Hietaniemi wrote:
Jarkko Hietaniemi wrote:
cc: Warning: perl.h, line 4175: In the initializer for
PL_vtbl_arylen.svt_get, the referenced type of the pointer value
Perl_magic_getarylen is function (pointer to struct sv, pointer to
const struct magic) returning int, which is not compatible
A patch for this is attached.
Uh-oh. A working patch is attached in here.
First time I've seen a zero-length file described as a working
patch :-(
Just rotate the bits 90 degrees clockwise, shake well, and look against
a strong light. Oh, right, there are no bits to rotate. Try these
cc: Warning: perl.h, line 4175: In the initializer for
PL_vtbl_arylen.svt_get, the referenced type of the pointer value
Perl_magic_getarylen is function (pointer to struct sv, pointer to
const struct magic) returning int, which is not compatible with
function (pointer to struct sv, pointer to
The information about the cross-compilation was a little bit dusty.
--- INSTALL.dist2005-06-30 08:14:58.0 +0300
+++ INSTALL 2005-06-30 08:36:49.0 +0300
@@ -1859,36 +1859,75 @@
=head2 Cross-compilation
-Starting from version 5.8, Perl has the beginnings of
...
cd t
LD_LIBRARY_PATH=/cluster/members/member0/tmp/jhi/perl-5.9.x:/tmp/jhi/perl-5.9.x
./perl TEST -deparse /dev/tty
--
TESTING DEPARSER
--
Tru64 picky cc gripes about a test in bleadperl:
lib/ExtUtils/ParseXS/t/basic..
cc: Warning: XSTest.xs, line 117: The scalar variable i is fetched but
not initialized. And there may be other such fetches of this variable
that have not been reported in this compilation. (uninit1)
Tom Horsley wrote:
temp slot), by doing a double cast via an (unsigned) integer.
I hope that means unsigned long, not unsigned int, otherwise
I write 'integer', you read 'int'. Most curious :-)
on lots of 64 bit platforms you'll truncate pointer values to 32
bits :-)
The attached patch does data pointer - function pointer casting
a bit differently than my earlier attempts at it, which tunneled
the pointers through a union. There's an easier way to do it which
feels slightly less unclean (and avoids assigning the value to a
temp slot), by doing a double cast
Gisle Aas via RT wrote:
I suggest we make chr($n) return \x{FFFD} when $n is 0.
Ahh, yes, that's one possibility, see e.g.
http://www.unicode.org/glossary/
Replacement Character. Character used as a substitute for an
uninterpretable character from another encoding. The Unicode Standard
uses
Nicholas Clark via RT wrote:
On Sun, Jun 05, 2005 at 05:27:30PM -0400, Ed Allen Smith wrote:
Under use bytes, isn't that the case? If it isn't, under what pragma or
whatever can one indeed get a guarantee of no characters outside ASCII? (If
the answer is there isn't one, then I suggest this
This time both a speedup and a cleanup: I brought the inlined utf8
goodness to is_utf8_string_loc(), too, but while doing that noticed
that at one place in mg.c is_utf8_string() was being used wastingly
so that the subcaptured string was walked through twice, first for
utf8 validity and then for
The attached patch allows one to build perl.pixie [1] even with
optimized *and* debug-symboled Perl (-Doptimize='-O4 -g3'), not
just with debug-symboled (-Doptimize='-g'). The patch affects
only Tru64.
[1] compiler backend that produces basic block instrumented Perl,
useful for profiling
Hmm. pre-5.6.0 let you use signed characters (e.g. chr(-3) is
chr(253)), and 5.6.0 and later still do under use bytes. I'd hate
Sigh. Backward compatibility is such a pain. And since there is
no fuller definition of the language than the implementation, anything
unspecified and undefined
Ed Allen Smith wrote:
In message [EMAIL PROTECTED] (on 6 June 2005 00:07:49 +0300),
[EMAIL PROTECTED] (Jarkko Hietaniemi) wrote:
it's the concept of negative input to chr() that bothers me, a lot,
since it assumes a fixed and closed range of characters,
Under use bytes, isn't
cc: Warning: sv.c, line 4959: In the initializer for pvx, the referenced
type of the pointer value ((const char ...)((sv)-sv_u.svu_pv)) is
const, but the referenced type of the target of this assignment is not.
(notconstqual)
char *pvx = SvPVX_const(sv);
^
I further inlined the fast UTF8 path in is_utf8_string(),
thus cutting down the number of function calls in a tight loop.
With this the run time of t/uni/class.t was cut by further
20-25% (relative to the first speed up), so now the cumulative
total speedup is in the category of 30-40% (e.g. in
I got annoyed by t/uni/class.t being so slow, so I inlined
much of the is_utf8_char(). This seems to cut away about 25%
of the run time of class.t (both in x86/linux and tru64).
Still slow, though.
utf8.pat.bz2
Description: Binary data
Based on Craig's suggestions please find updated patches,
and a new one for perl.h.
pat.tgz
Description: GNU Zip compressed data
Jarkko Hietaniemi wrote:
Based on Craig's suggestions please find updated patches,
and a new one for perl.h.
It would be good if the updated patches were actually the
updated ones, no?
pat.tgz
Description: GNU Zip compressed data
the dl_open.xs.
--
[EMAIL PROTECTED] Jarkko Hietaniemi http://www.iki.fi/ I wish life was not
so short, he thought. Languages take such a time, and so do all the
things one wants to know about. -- J. R. R. Tolkien, The Lost Road
--- sv.c.dist 2005-05-29 18:30:10.0 +0300
+++ sv.c2005
Jarkko Hietaniemi wrote:
I think nobody has tried t/uni/class.t with -w for a while, there are
quite a lot of warnings.
Here's a patch based on Sadahiro-san's suggestions:
--- t/uni/class.t.dist 2005-05-28 11:20:50.0 +0300
+++ t/uni/class.t 2005-05-28 11:41:19.0 +0300
Nicholas Clark wrote:
On Thu, May 26, 2005 at 08:36:19AM +0300, Jarkko Hietaniemi wrote:
There is no lchown() in MPE/iX, as found out by Donna Garverick
and analyzed by Mark Bixby. Suitable for both blead and maint
(I would say even the 5.8.7 RC1, because otherwise the POSIX extension
ends up
I think nobody has tried t/uni/class.t with -w for a while, there are
quite a lot of warnings.
There is no lchown() in MPE/iX, as found out by Donna Garverick
and analyzed by Mark Bixby. Suitable for both blead and maint
(I would say even the 5.8.7 RC1, because otherwise the POSIX extension
ends up broken in MPE/iX, which makes many tests emit black smoke.)
--- hints/mpeix.sh.dist
Thanks, applied in its present minimal form.
You're probably right about using the compiler as the linker as the
default as a good idea, simply because I'm sure you know more about
linking than I do. But I feel like the conversation may not have
I don't know about that... but at least
Two of them, same cause:
cc: Warning: perly.c, line 325: In this statement, the referenced type
of the pointer value ((const char ...)((yyns_sv)-sv_u.sv_pv)) is
const char, which is not compatible with pointer to const char.
(ptrmismatch)
yyns = SvPVX_const(yyns_sv);
cc: Warning:
if
the executable ever wants to call shared libraries. Similar things
are needed in many systems. I think in Linux the GNU ld just hides
much of these things so that they are not visible, even if using
the command line -v option or somesuch.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi
Obviously Configure knows this though...perhaps some kind soul can
It's not really Configure as such. It's a combination of the hints
files info (which does get saved as %Config, but I'm not entirely
certain that the naming of the fields is consistent across platforms,
e.g. which flags get
platforms, actually,
and especially so if C++ is involved. In fact I would go as far as
to claim that using the compiler as the linker is a better default
than using the linker.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ This principle is
so perfectly general that no particular
ok 1
Can't create t/compilet.c: No such file or directory at
../lib/ExtUtils/CBuilder/t/02-link.t line 39.
You need to set PERL_CORE; the test expects a different directory
structure when in the core vs. the standalone module; I'd recommend
doing so by one of:
./perl harness -v
Gisle, your t/run/exit.t sub run fix for HP-UX worked also for Tru64.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ I have always
wished that my computer would be as easy to use as my telephone.
My wish has come true. I no longer know how to use my telephone.
-- Bjarne Stroustrup
Tru64 is unhappy with [EMAIL PROTECTED]:
Failed Test Stat Wstat Total Fail Failed List of
Failed
---
../lib/ExtUtils/CBuilder/t/02-lin51 20.00% 5
run/exit.t
#define WNOHANG 0x1
#define WUNTRACED 0x2
#define _WSTOPPED 0177
#ifdef _BSD
#define _W_INT(w) (*(int *)(w)) /* convert union wait to int */
#else
#define _W_INT(i) (i)
#endif
#define _WSTATUS(x) (_W_INT(x) _WSTOPPED)
#if ( !defined(_OSF_SOURCE)
Gisle Aas wrote:
Jarkko Hietaniemi [EMAIL PROTECTED] writes:
Tru64 is unhappy with [EMAIL PROTECTED]:
Failed Test Stat Wstat Total Fail Failed List of
Failed
---
../lib/ExtUtils/CBuilder/t/02
Because the intention of the patch was exactly that you should be able
to assume the traditional UNIX/POSIX semantics for $?.
That's a bold claim given that e.g. signals, or a process being stopped,
or the concept of core dump, may or may not be supported in non-UNIX
(*) operating
Until Andy gets Configure __attribute__-aware (and even then
something like this could be useful for the places that have
gcc but no Configure).
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ Disgusting, no?
But it compiles and runs just fine. I feel a combination of pride
Jarkko Hietaniemi wrote:
The attached patch turns the (C89) strictness knob to eleven in Tru64
by changing the -std switch to -std1. The change doesn't turn up
anything
deadly (like C++ style use of // or variable declarations after code,
the kinds
of things that often make grumpiness
main
goal anyway.
(*) somewhat curiously the same technique has been used in
conjunction with
the xpvio type already earlier, search for xio_dirpu in sv.h.
std1.pat.bz2
Description: Binary data
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist
.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
On May 6, 2005, at 0.46, Alex wrote:
According to the perlrun manpage, setting PERL_UNICODE to 0 in the
environment or using the command line switch -C0 is supposed to
turn off all unicode processing. However I'm noticing that when I use
It doesn't say that. It says:
You can use -C0 (or 0
It would be interesting to see how Perl fares:
http://www.dwheeler.com/sloccount/
http://sourceforge.net/projects/
could be useful in pinpointing problem areas in the code.
, a new patch attached.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
--- Time-HiRes-1.66/HiRes.xs2004-12-19 22:10:53.0 +0200
+++ Time-HiRes-1.67tobe/HiRes.xs2005-05-03 09:00
.)
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
--- Time-HiRes-1.66/HiRes.xs2004-12-19 22:10:53.0 +0200
+++ Time-HiRes-1.67tobe/HiRes.xs2005-05-02 08:32:23.0 +0300
.)
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
--- Time-HiRes-1.66/HiRes.xs2004-12-19 22:10:53.0 +0200
+++ Time-HiRes-1.67tobe/HiRes.xs2005-05-02 08:32:23.0 +0300
Andy Lester wrote:
Two unused vars in Time::HiRes. I assume most compilers will optimize
it all away anyway, but here they are:
(resending since my email died)
Instead of throwing away information let's use it.
(See attached patch.)
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi
started to hurt. (The idea being
that the function pointer table sv_stre should be const and the
function pointers in it should be const.)
[1] [EMAIL PROTECTED]
[2] [EMAIL PROTECTED]
[3] [EMAIL PROTECTED]
Also testing in Win32 would be good.
So here's the patch.
--
Jarkko Hietaniemi [EMAIL
Here's an updated version: the patterns tweaked to gobble less
(especially in the gcc case).
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
grepccerr.bz2
Description: Binary data
/solaris
owners are welcome to play with the tool (attached).
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
Err-24082.bz2
Description: Binary data
grepccerr.bz2
Description: Binary data
if Abe would consider integrating grepccerr into Test::Smoke
so that if the build fails at the M or m stage, the reason is there for
everyone to see in the smoke report.
I wonder that, too. wink wink nudge nudge
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
A number of consting mismatches, see the attachment.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
Warning.bz2
Description: Binary data
happened.
Is this a terribly bad thing?
If there's a way to rewrite New() not to multievaluate its argument,
that would be good. If not or even if not, grepping the code for
similar problem spots would be not be bad.
Nicholas Clark
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi
not solved by patch 23941)
Nicholas Clark
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
perl-current compiled with -g
valgrind --num-callers ./perl -e '$^V'
gives Invalid read of size 1 ... 0 bytes after a block of size 6,
report attached.
The suggested cure: make Perl_savesvpv() more like Perl_savepvn().
Patch attached.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi
.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
according to the alloc trail.
Nicholas Clark
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
the cell - Perl core code has much too
often too much understanding for bad input, even Postel's law can
overapplied...)
How about making (in the blead) the code to have an assert() in
savesvpv() for detecting such non-\0-terminated strings?
Nicholas Clark
--
Jarkko Hietaniemi [EMAIL PROTECTED
than was meant. Easily reproducible in older
Perls before malloc wrapping with the autoextension of Perl
arrays, for example.
IRIX problems with it.
Thanks!
-Allen
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use
MG_VIRTUAL = PL_vtbl_utf8
MG_TYPE = PERL_MAGIC_utf8(w)
MG_LEN = 2
looks like the SvLEN of the PADTMP associated with the uc op is doubling
each second time round the loop.
Over to you, Jarkko ???
Urque.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi
Nicholas Clark wrote:
On Sun, Jan 23, 2005 at 07:52:57PM +0200, Jarkko Hietaniemi wrote:
Over to you, Jarkko ???
Urque.
Hmm... I can't seem to find a patch in there anywhere.
Finding a patch there would have violated several known laws of physics,
most importantly it would have
either it's working
perfectly or it hasn't been downloaded and tested by
enough people, in which case putting it on CPAN would
get more eyeballs on it.
Cheers,
-Marc
=
Marc Abramowitz - (H) 408.248.3280 (W) 408.349.7412 (M) 650.387.5430
http://marc.abramowitz.info
--
Jarkko Hietaniemi
Dave Mitchell wrote:
On Sun, Jan 23, 2005 at 09:25:01PM +0200, Jarkko Hietaniemi wrote:
Finding a patch there would have violated several known laws of physics,
most importantly it would have involved time travel.
Umm, much as I hate being the tider of bad news, with that patch
Since dropping the optimisation level of gcc seems to fix the failures,
here's a patch to that effect. Pity that we have to go all the way down
to -O1 from -O3. Maybe some future gcc release will be better.
Nicholas Clark
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi
. The SvGROW()s are better done on the SvLEN()s.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
case2.pat.gz
Description: GNU Zip compressed data
Ed Allen Smith wrote:
In message [EMAIL PROTECTED] (on 22 January 2005 19:45:02 +0200),
[EMAIL PROTECTED] (Jarkko Hietaniemi) wrote:
Since dropping the optimisation level of gcc seems to fix the failures,
here's a patch to that effect. Pity that we have to go all the way down
to -O1 from -O3
the naughty assumption.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
packsize.pat.gz
Description: GNU Zip compressed data
Looking further, the change in question is one of Jarkko's (see below)
Jarkko, does this ring any bells?
Trying to unring some bells.
Dave.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack
[EMAIL PROTECTED] wrote:
Jarkko Hietaniemi [EMAIL PROTECTED] wrote:
:Of course if we want to be sneaky we could just malloc the same
:number of bytes for the result as there are in the input and then
:in the loop do any necessary reallocing, since there are after all only
:63 Unicode
[EMAIL PROTECTED] wrote:
Jarkko Hietaniemi [EMAIL PROTECTED] wrote:
:Of course if we want to be sneaky we could just malloc the same
:number of bytes for the result as there are in the input and then
:in the loop do any necessary reallocing, since there are after all only
:63 Unicode
. But then
someone will no doubt try to casefold a gigabyte of U+0390's and
complain about the excessive number of reallocs...
Dave.
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
]
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
for POSIX
--
Jarkko Hietaniemi [EMAIL PROTECTED] http://www.iki.fi/jhi/ There is this
special
biologist word we use for 'stable'. It is 'dead'. -- Jack Cohen
1 - 100 of 439 matches
Mail list logo
