at least in Solaris 8/SPARC, the
system headers won't work with -ansi -pedantic. I'd be surprised if
things were any different on Solaris 10/x86.
--
Andy Dougherty [EMAIL PROTECTED]
troduced after Configure used 'make'
I can think of two possibilities offhand: Does the user have an
environment variable set that might lead to this? Alternatively, did
the previously-installed perl (/usr/bin/perl) use gmake?
I could imagine it's possible that the Constant.t test gets confused about
which $Config{} to load up. Obviously, that would be a bug.
--
Andy Dougherty [EMAIL PROTECTED]
't beat me to it, I'll try to look at it in the next couple
of days.
--
Andy Dougherty [EMAIL PROTECTED]
wly-installed perl is first in your $PATH, then it
ought to find it just fine. Failing that, you could temporarily rename
the faulty /opt/tools/bin/perl5 file to something else and then re-run
Mozilla's configure.
--
Andy Dougherty [EMAIL PROTECTED]
glong=define, longlongsize=8, d_longdbl=define, longdblsize=16
>ivtype='long', ivsize=4, nvtype='double', nvsize=8, Off_t='off_t',
> lseeksize=8
>alignbytes=8, prototype=define
> Linker and Libraries:
> ld='ld', ldflags =' '
>libpth=/localdisk/vahik/05septBuild/lib
>libs=
>perllibs=
>libc=/lib/libc.so, so=so, useshrplib=false, libperl=libperl.a
>gnulibc_version=''
> Dynamic Linking:
>dlsrc=dl_none.xs, dlext=none, d_dlsymun=undef, ccdlflags=''
>cccdlflags='', lddlflags=''
Hope this helps,
--
Andy Dougherty [EMAIL PROTECTED]
Using
this program the script would be setuid and look like this:
#!/bin/indir -u
#?/bin/sh /safe/path/to/this/script
/bin/rm /some/directory/lockfile
> Does anyone know?
I think it's probably safe to drop that part from ponie :-).
> Nicholas Clark
--
Andy Dougherty [EMAIL PROTECTED]
t-in specs' in gcc-4. Perhaps we should just use the same
> > gcc test as +# in Configure to see if we're using gcc.
> > +if echo "$verbose" | egrep '(Reading specs from)|(Using built-in specs)'
> > >/dev/null 2>&1; then #
> > # Using gcc.
> > #
>
> Is this patch still to be applied?
Yes, please. We've now verified it works.
--
Andy Dougherty [EMAIL PROTECTED]
od idea to set the archname to
something appropriate. Setting -Darchname should do the trick. (You can
ignore the targetarch stuff since you're not doing a cross-compile.)
> Configure was not expecting those variables to be defined from the command
> line. What you should do is create your own Policy.sh, and keep that updated
> to your needs.
That will also work.
--
Andy Dougherty [EMAIL PROTECTED]
On Tue, 26 Jul 2005, Niki Waibel wrote:
> On 25-Jul-2005 Andy Dougherty via RT wrote:
> > On Thu, 10 Jun 2005, Niki W. Waibel wrote:
> >
> >> # New Ticket Created by "Niki W. Waibel"
> >> # Please include the string: [perl #36235]
> >> # in
gcc-4.0 -v -o hello hello.c > gcc4.out 2>&1
gcc-4.0 -Wl,-V -o hello hello.c > gcc4-Wl.out 2>&1
(adjusting the exact gcc names as needed.)
and send me the various *.out files? Then perhaps I can see
why the hints/solaris_2.sh file is failing.
--
Andy Dougherty [EMAIL PROTECTED]
ss-compilation
and locales, and doing something clean and portable, like signals :-).
Cheers,
Andy Dougherty [EMAIL PROTECTED]
ompiled.)
My initial suspicion is something in your environment, perhaps related
to /usr/ucb, or perhaps related to the type of filesystem on which you
are running Configure.
--
Andy Dougherty [EMAIL PROTECTED]
On Wed, 22 Jun 2005, Andy Dougherty wrote:
> On Wed, 22 Jun 2005, Paul Johnson wrote:
>
> > On Tue, Jun 21, 2005 at 07:50:35PM +0200, Salvador Fandino wrote:
> >
> > > I have not been able to reproduce the bug on any perl compiled by
> > > myself, but anyway
ler.
That'd be me. What, precisely, would you like me to test? I haven't had
any recent problems with that compiler under -O, so as far as I know, but
I haven't compiled bleadperl for a few weeks.
--
Andy Dougherty [EMAIL PROTECTED]
to point out that
> perl5.6.1 gives the same results as perl5.8.7 so it's been deprecated
> long enough.
I would certainly assume it won't get rolled into maint. It'll break code
that's currently working. Hence Nick's "Gah" response.
--
Andy Dougherty [EMAIL PROTECTED]
ck them up.
(The main "offenders" are extensions.)
Further, if you try LEAKTEST, you'll also end up calling
Safefree, which might call safexfree() on some things that weren't
malloced with safexmalloc. The correct "fix" to this, if anyone
is intere
nd delete all the v8plus entries.
(Actually, you can probably delete all the
-L/opt/Studio10/SUNWspro entries.)
3. sh Configure -S
4. make depend
5. make
--
Andy Dougherty [EMAIL PROTECTED]
On Tue, 24 May 2005, Nicholas Clark wrote:
> On Tue, May 24, 2005 at 01:11:32PM -0400, Andy Dougherty wrote:
> > On Tue, 24 May 2005, Nicholas Clark wrote:
> >
> > > On Tue, May 24, 2005 at 02:27:15PM +0100, Steve Hay wrote:
> > > > Nicholas Clark wrote:
&g
straint on NVs is greater than IVs, then
> the latter will have padding.
That's not an uncommon constraint. SPARC works the same way.
If you want to do things conditionally, Configure probes for NV alignment.
It's available as $Config{alignbytes} and MEM_ALIGNBYTES in config.h.
--
Andy Dougherty [EMAIL PROTECTED]
On Mon, 23 May 2005, Nicholas Clark wrote:
> On Wed, May 18, 2005 at 12:51:37PM -0400, Andy Dougherty wrote:
>
> > Hmm. Looking back a little in my archives -- it appears that the
> > "STRANGE_MALLOC" was something I introduced in perl5.002b1h to work around
>
ous that has changed in perl that would
cause this problem. Might something have changed on your system? It
might be a good idea to try rebuilding perl 5.8.5 again to see if it
succeeds or if it fails in a similar way.
--
Andy Dougherty [EMAIL PROTECTED]
e, Ilya improved perl's malloc so
that it didn't have this problem, and hence he added the || MYMALLOC
to the STRANGE_MALLOC case.
In any case, if you're going to change that part, make sure to go into
hv.c and av.c and check out the #ifdef STRANGE_MALLOC cases -- I think you
want them enabled in all cases now.
--
Andy Dougherty [EMAIL PROTECTED]
; 2. Kernel 2.6.11.3
> 3. Binutils 2.14.24.0.2
> 3. make 3.8.0
> 4. gcc c and c++ compiler 3.4.3
The output of the ./myconfig script is sometimes helpful in debugging such
cases too. Please include that if you follow up on this.
--
Andy Dougherty [EMAIL PROTECTED]
should be able to
run
PERL_CORE=1 ./perl -Ilib lib/Net/hostent.t
to see more details of the failing tests.
--
Andy Dougherty [EMAIL PROTECTED]
lot of the surrounding context
to help understand all the other changes you've proposed. Then,
throughout the build patches, a very healthy dose of comments
understandable even by those who don't even know what NetWare is would
probably be very helpful.
--
Andy Dougherty [EMAIL PROTECTED]
s.
The best way to debug it is by hand -- copy the exact program Configure
tried to use into a file 'try.c' and then try to compile it yourself with
the exact command line Configure used. Vary it around until you get
something that works. Once you do, please report back to perlbug so we
can help others avoid the same problem in the future.
--
Andy Dougherty [EMAIL PROTECTED]
uild. I don't
currently have time to wade through it all, but I suspect most of the
complaints are either incorrect or very very hard to get rid of.
Have fun,
--
Andy Dougherty [EMAIL PROTECTED]
bleadperl-lint.txt.bz2
Description: Lint output
6.
For more information, browse http://www.sun.com/developer-products/
--
Andy Dougherty [EMAIL PROTECTED]
using 'const', since various operating systems
have traditionally been wildly inconsistent about whether or not various
functions take 'const'.
If the data stays within perl, then no, offhand I can't think of any
reason to avoid 'const'.
--
Andy Dougherty [EMAIL PROTECTED]
}
- goto unknown;
-
case 'k':
{ /* readlink */
return -KEY_readlink;
}
-
- goto unknown;
default:
goto unknown;
--
Andy Dougherty [EMAIL PROTECTED]
case, Pod::Usage probably ought to either check the return
code from system() or check that $progpath actually exists before blindly
calling it.
--
Andy Dougherty [EMAIL PROTECTED]
On Mon, 14 Feb 2005, Marcus Holland-Moritz wrote:
On 2005-02-14, at 12:06:09 -0500, Andy Dougherty wrote:
Note the extra 'goto unknown' sprinkled in between each 'case'
statement.
You're right. The statements don't hurt, but they're indeed superfluous.
I'v
/* q */
return KEY_q;
}
goto unknown;
--
Andy Dougherty [EMAIL PROTECTED]
ll, it might be on your system, but it's certainly not everywhere.
Your broader point is correct -- Configure in many little ways still shows
its BSDish origins. However, it generally works, so there's little
incentive to muck with things that work :-).
--
Andy Dougherty [EMAIL PROTECTED]
t because Configure
was designed for Unix-like systems, and such systems have historically
always had a /bin/echo command, though different systems have slightly
differing versions.
The problem is to figure out why Configure couldn't figure out how to
use your /bin/echo command. I'll need to know more about whatever
'msys' is and what, exactly, you had that fooled Configure into
thinking there was a /bin/echo command.
--
Andy Dougherty [EMAIL PROTECTED]
rong values. I do note,
however, that you ran Configure interactively (which is certainly
fine) so it's possible you accidentally typed something wrong in and
Configure didn't give you helpful enough feedback about it.
If you wish to pursue it, start with a clean directory and try
re-running Configure from scratch. If you just want to get on with
building perl, then the config.sh changes I list above are probably
all that is needed.
Hope this helps,
--
Andy Dougherty [EMAIL PROTECTED]
also seems like the right thing to do for various
> reasons:
Oh, I agree. I'm just not volunteering to do it!
--
Andy Dougherty [EMAIL PROTECTED]
'd be
something like
libc='/lib/libc-2.3.2.so'
(Configure doesn't actually use that information anyway, but it's
possible that whatever tripped up Configure in finding your C library
prevented it from suggesting the -lpthread library as well. Or
perhaps something else went amiss.)
In any case, you need to go back to the beginning and re-run
Configure, this time including your pthreads library.
Hope this helps,
--
Andy Dougherty [EMAIL PROTECTED]
gt; > behave in the same way as dynamic_ext w.r.t. Thread. Is that correct?
Yes. In Configure, it is simply constructed by shell code equivalent to
extensions="$dynamic_ext $static_ext $nonxs_ext"
--
Andy Dougherty [EMAIL PROTECTED]
On Tue, 18 Jan 2005, Jos I. Boumans wrote:
>
> On Jan 18, 2005, at 7:20 PM, Andy Dougherty wrote:
> >> All this desribes the bigger problem of: What do we do when we want to
> >> integrate module into the core that require core XS modules to build
> >> themselves?
time with just the "simple" extensions, and the second time with
all the extensions.
I don't see any fundamental problem with such a re-organization, but it
does look a little tedious to implement.
--
Andy Dougherty [EMAIL PROTECTED]
I've never run into this problem, so I can only state the obvious.
How essential, really, is FileHandle to the Makefile.PL? What is
Makefile.PL doing that is so complex? (Note that I've never installed
CPANPLUS, so I have no idea what's going on.!)
--
Andy Dougherty [EMAIL PROTECTED]
Another approach sometimes taken for this whole issue (as I recall) was to
edit the binaries at install time. That would handle the LD_RUN_PATH
stuff, but, of course, requires the user to run a special program instead
of a plain 'mv' to move the perl installation. That constraint might well
be one you were hoping to avoid.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
that script.
Do other 'fbsd' (which I'll guess is FreeBSD) webmin users see the same
problem, or is it apparently working for them? This also might help pin
down whether it is something specific to your system (e.g. locale
settings, other installed packages, or some such).
--
Andy Dougherty [EMAIL PROTECTED]
On Thu, 16 Dec 2004, Michael G Schwern wrote:
> On Thu, Dec 16, 2004 at 09:57:44AM -0500, Andy Dougherty wrote:
> > > If POSIX was loaded and works, it should be quiet about it.
> > > If it was loaded but doesn't work (ie. its autoloaded functions are non
> > >
2002-03-19 17:14:33.0 -0500
+++ perl-andy/ext/POSIX/Makefile.PL 2004-12-16 09:40:34.0 -0500
@@ -1,3 +1,7 @@
+# Explicitly avoid including '.' in @INC; autoloader gets confused since it
+# can find POSIX.pm, but can't find autosplit.ix.
+BEGIN { @INC = '.
ODBM_File if OSR5 because it's mostly broken
# but keep it for ODT3 as it seems to work.
if test "$scorls" = "5"; then
--
Andy Dougherty [EMAIL PROTECTED]
;;
> *)
> -ccdlflags='-Bexport -L/usr/local/lib'
> +ccdlflags='-Wl,-Bexport -L/usr/local/lib'
> cccdlflags='-Kpic'
> lddlflags='-G -L/usr/local/lib'
> ;;
>
>
--
Andy Dougherty [EMAIL PROTECTED]
.8.6 might fix
it, but it's hard to see how you'd hit that limit if no array size is over
25.
I have a number of versions of perl running on Solaris 8. If you could
send me (Privately, OFF LIST) a copy of a *working* script that crashes
along with an appropriate data file, I could see if it's reproducible.
--
Andy Dougherty [EMAIL PROTECTED]
in question is
int Perl_getcwd_sv(pTHX_ register SV *sv)
The problem is the pTHX_. The original poster is trying to build this
with perl5.005_03, which doesn't have the pTHX_ defines.
The others errors are probably similar, due to stuff not available in the
old 5.005_03 package. I
00,'very basic times test');
($sec,$min,$hour,$mday,$mon,$year,$wday,$yday,$isdst) = localtime($beg);
($xsec,$foo) = localtime($now);
--
Andy Dougherty [EMAIL PROTECTED]
d us the output of installperl prior to
where it got the error message? It might help us track it down.
--
Andy Dougherty [EMAIL PROTECTED]
lib/Carp/Heavy.pm file?
If it still fails, please send us the full ./myconfig output and maybe
someone else will be able to reproduce the problem.
--
Andy Dougherty [EMAIL PROTECTED]
p5p archives somewhere.)
To some extent, however, this problem may be rarely, if ever, encountered.
Configure's list is generated by a C program linked with the same set of
libraries as perl will be linked, so it's generally going to get the right
answers.
--
Andy Dougherty [EMAIL PROTECTED]
ome library or other, or you need to undefine some feature
+that Configure thought was there but is defective or incomplete. If
+you used a hint file, see if it has any relevant advice. You can also
+look through through config.h for likely suspects.
=item toke.c
--
Andy Dougherty [EMAIL PROTECTED]
On Mon, 25 Oct 2004, Jim Cromie wrote:
> in http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2004-09/msg00393.html
>
> Andy Dougherty asked:
>
> I ran make test on a freshly-checked-out 5.9.x today (.patch = 23334) and
> test/harness reported that tests UNEXPECTEDLY S
ssful (4 subtests UNEXPECTEDLY SUCCEEDED), 52 tests and
487 subtests skipped.
--
Andy Dougherty [EMAIL PROTECTED]
onfigure script. It doesn't normally use
environment variables to set things -- instead, you define things with -D
options. In particular, Configure itself won't do anything with
those CC, LD, or AS environment settings.
--
Andy Dougherty [EMAIL PROTECTED]
Please see the section
=head2 Cross-compilation
in the INSTALL file.
--
Andy Dougherty [EMAIL PROTECTED]
he same as what you see, but the closest thing
I can think of.
Further, your error message is preceded by a './perl:' token.
Perl generally doesn't include it's name in the error message,
so I don't think that message is really coming from perl.
--
Andy Dougherty [EMAIL PROTECTED]
On Thu, 20 Nov 2003, Daniel Berger wrote:
> > From: Andy Dougherty [mailto:[EMAIL PROTECTED]
> > I don't have gcc-3.3 to test (my Solaris gcc installation is
> > still 2.8.1), but I think the following patch should fix it:
> >
> > --- perl-5.8.2/hints/sola
we used the '&'. I'm
working on writing down this collected wisdom so we have a place to look
the next time gcc next breaks this test.
--
Andy Dougherty [EMAIL PROTECTED]
ere is no '0' at the end of "Makefile"
Ok, I'm pretty sure that's a makedepend issue then. I'll have a look at
how we might go about debugging it.
One other thought -- have you tried with AIX make instead of GNU make?
--
Andy Dougherty [EMAIL PROTECTED]
def'. How was the config.sh you
included with your perlbug report generated? Did perlbug foul that up on
you, or is something else at work?)
In short, I vaguely suspect some sort of conflicting previous
installation of libperl, but I can't figure out your particular set-up
well
ns ") echo "Ignoring extension $i." ;;
+case " $noextensions " in
+*" $i "*) echo "Ignoring extension $i." ;;
*) echo "Keeping extension $i.";
keepextensions="$keepextensions $i"
;;
--
Andy Dougherty [EMAIL PROTECTED]
to test, so I probably can't be much
further help here.
--
Andy Dougherty [EMAIL PROTECTED]
alue of SRCHTIM is.
This is actually a bug in your printf statement. In order to print a raw
"%", you need to use "%%". That is, you should use
printf ("%0.2f\n",$SRCHTIM);
That said, your form did do what you expected with perl 5.6, but I expect
th
us about this. I'll sit on that one for the moment.
It should be ok as long as you use a copy (or a wrapper, though that'd
might be a bit silly in this case) There is no really portable way to
create a link (hard or soft).
--
Andy Dougherty [EMAIL PROTECTED]
l not be binary compatible with 5.6.1, so those
issues don't exist in quite the same way. (They will for 5.8.1, though,
so I didn't want to delete that whole section.)
I'm currently updating to perl@17623 and will try to look at your version
and make specific suggestions. (Quickly,
s over in ext/POSIX/hints/svr4.pl -- still mostly
a cut-and-paste operation, I think. [he says, without either cutting or
pasting anything himself . . . ]
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
On Mon, 10 Jun 2002, Nicholas Clark wrote:
> On Mon, Jun 10, 2002 at 04:22:42PM -0400, Andy Dougherty wrote:
>
> > I'm sorry you had trouble, but I'm unsure if a warning would have helped.
> > Would you have seen such a warning fly by in the Configure dialog?
&
0,
Have you updated your kernel headers recently such that there's possibly a
mismatch between your libc headers and the kernel ones?
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
,-E"
- lddlflags="$lddlflags -W,l-E -G"
+ lddlflags="$lddlflags -Wl,-E -G"
fi
fi
--
Andy Dougherty [EMAIL PROTECTED]
t; > Bus error
This error is now gone for me with perl@13016. I can't figure out why,
but it makes me happy.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
On Fri, 26 Oct 2001, Jarkko Hietaniemi wrote:
> On Thu, Oct 25, 2001 at 09:00:16PM -0400, Andy Dougherty wrote:
> > >I think the 'minitest' is not a 'normal' target to try out 'first',
> > >it's only to be used when the real build has fail
's/
+$//'`
# This is strange.
usevfork="$undef"
esac
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
ink of is something like AFS or DFS.
> Is there anyway to explicitly check what sort of filesystem you're on,
> instead of implying it from $^O?
Sadly, no.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
2. Use the vendorlib hierarchy of directories for vendor-specific
modules. These use the same version-searching algorithm that the site-lib
directories do.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
f however you take the (weighted) average runtime of the tests (the only
> right way to combine benchmark scores), you will see perl slowed down
> 20%-40%).
Well, the perl test suite spends a lot of time sleep(3)-ing these days,
sot those timings can also be deceiving.
--
Andy D
n Mac OS X. Who knows what vendor files will be getting
included and parsed, and with -traditional, gcc will not define
__STDC__ and may thus take a slightly different path through all system
include files.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
one. You may have to manually add in either
/vol/SUNWspro/WS6/lib/v9 or /opt/SUNWspro/lib/v9 prior to the regular
SUNWspro libraries, depending on which installation you wish to use.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
onfigure ended up missing them, but if you can reproduce
the problem with a fresh run, please let us know how Configure messed up.
Good luck,
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
amples in INSTALL.)
In any case, glad to hear it's working now.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
is that regen_headers ought to be done before ./perl
is even built. Hence the person running it must have a 'perl' elsewhere
in $PATH. Normally this isn't a problem since normally only a 'porter has
to run regen_headers
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
.
> > Do we really *need* more heading levels? What for?
and I replied:
> No, we don't really *need* more heading levels. However, I do *want*
> them. [example deleted.]
Just to be clear and explicit: Russ is the current author for Pod::Man.
If he still objects, we should resp
e quick, however, and I begin to wonder if the
perl6 tail ought to wag this particular dog.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
l that was really missing was the Configure hook to get installperl to
operate in versiononly mode.
This patch changes the installscripts hook to 'versiononly', and it now
maps exactly onto installperl's versiononly variable.
--
Andy Dougherty [EMAIL PROTECTED]
on.
I'm just not sure how hard it's worth working on the MakeMaker end to deal
with this sort of issue. I'm hoping someone will have an elegant insight
and implement something much nicer.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
refuse to apply this to the official
> >sources. I understand the need but this needs to be solved somehow
> >better.
I will timidly admit that I've been (very slowly) working on this. I hope
to report back with something by the end of next week.
--
Andy Dougherty
want to remove the libraries at the end of Configure. That
will foul up folks who re-use config.sh (either intentionally or
unintentionally). See
http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/1999-06/msg00072.html
for more details and a possible different solution.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
porary files. There's no way we can reliably test whether all
compilers do so securely and, if not, whether we can stop all of them from
doing so.
So don't run Configure/make as root.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easto
/**/
and Configure will look for mkstemp().
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
the user to override the defaults and blend these separate
installations, but I wouldn't want that to be the default.
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
to use.
Still, I gather this is all interim stuff, as I imagine 5.6 won't really
have 2 different supported "threads" models. So I don't really want
to work too hard on the problem.
> If you have tuits to spare, please consider working on these.
Spare? Who has tuits to spa
94 matches
Mail list logo
