[perl #41195] [BUG]: Change to Configure.pl causing 'make' to fail on Darwin
James Keenan
# Please include the string: [perl #41195]
# in the subject line of all future correspondence about this issue.
# <URL: http://rt.perl.org/rt3/Ticket/Display.html?id=41195 >
Index: Configure.pl
===================================================================
--- Configure.pl (revision 16268)
+++ Configure.pl (revision 16416)
@@ -464,9 +464,9 @@
init::defaults
init::install
init::miniparrot
+ inter::progs
init::hints
init::headers
- inter::progs
inter::make
inter::lex
inter::yacc
Steve Peters via RT
James Keenan
On Jan 7, 2007, at 8:44 AM, Steve Peters via RT wrote:
> What is your c++ symlink pointing at?
>
>
>
[parrot] 512 $ ls -l /usr/bin/c++
lrwxr-xr-x 1 root wheel 7 Aug 9 2004 /usr/bin/c++ -> g++-3.3
[parrot] 513 $ ls -l /usr/bin/g++-3.3
-r-xr-xr-x 1 root wheel 135816 May 14 2006 /usr/bin/g++-3.3
Steve Peters via RT
OK, this problem must be something similar to the linking problems I
have with suncc on Linux. Even though I specify everything on the
Configure.pl command-line in terms of --cc and --link, it magically
switches to g++ when trying to link. Obviously, then, the hints or
something that runs after is misbehaving and redefining variables
configuration variables.
James Keenan
On Jan 11, 2007, at 9:21 AM, Steve Peters via RT wrote:
> On Sun Jan 07 08:27:28 2007, jk...@verizon.net wrote:
>>
> OK, this problem must be something similar to the linking problems I
> have with suncc on Linux. Even though I specify everything on the
> Configure.pl command-line in terms of --cc and --link, it magically
> switches to g++ when trying to link. Obviously, then, the hints or
> something that runs after is misbehaving and redefining variables
> configuration variables.
>
As noted previously, once I reverted to the previous order of running
the steps, I had no problem. I've successfully run 'make' quite a
few times since then.
Will Coleda via RT
> Two months ago, I filed this bug report (excerpt):
I'm still not seeing the effects of the bug you describe, btw. (on OS X
intel or ppc). It does remind me of issues I had when attempting to use
the GMP libs, though.
> > ...I inferred that some change in the configuration process in the
past 3
> > days had caused Configure.pl to override my environmental setting of
> > 'ld' to '/usr/bin/g++-3.3'.
> >
> > Running an 'svn diff' on the last two versions of Configure.pl (see
> > attached recent.changes.Configure.pl.txt), I saw:
> >
> > Index: Configure.pl
> > ==========================================
> > --- Configure.pl (revision 16268)
> > +++ Configure.pl (revision 16416)
> > @@ -464,9 +464,9 @@
> > init::defaults
> > init::install
> > init::miniparrot
> > + inter::progs
> > init::hints
> > init::headers
> > - inter::progs
> > inter::make
> > inter::lex
> > inter::yacc
> >
> > ... indicating that step 'inter::progs' is to be performed
> > immediately before 'init::hints', rather than two steps afterwards.
> > I peered into config/inter/progs.pm (for the very first time) and
> > noted that it has a provision for setting the value of 'ld'. config/
> > inter/progs.pm itself had not been changed in the last 3 days,
> > therefore I hypothesized that my problems were *solely* to the patch
> > applied to Configure.pl on Jan 04.
> >
> > Once I ran 'make realclean' and 'svn update', I then manually
> > reverted Configure.pl to have 'inter::progs' done two steps after
> > 'init::hints'. I then examined the resulting 'myconfig' and saw that,
> as I desired, line 12 once again ran:
> >
> > ld='/usr/bin/g++-3.3'
> >
> > I then ran 'make', which DWIMmed. ...
> >
> > On the basis of the above, I recommend that Configure.pl be reverted
> > to r16268. (If that cannot be done, then I would like to know how to
> > call Configure.pl so that I can 'make' Parrot successfully.)
> >
>
> The bug has not been addressed. Up until the last few days, I was
> content to let it go. But now I'm worried that it may be the cause of
> the bug t/pmc/object-mro.t I reported this morning
> (http://rt.perl.org/rt3/Ticket/Display.html?id=41704).
Do you still get the bug when you revert the change to Configure.pl?
> Thoughts?
>
> kid51
I tried reverting the change in a local copy of svn head today: seems to
work the same for me either way, so unless another platform objects,
let's switch it back if it fixes things for you.
James Keenan
On Mar 9, 2007, at 10:42 AM, Will Coleda via RT wrote:
> On Mon Mar 05 16:57:47 2007, jk...@verizon.net wrote:
>> Two months ago, I filed this bug report (excerpt):
>
> I'm still not seeing the effects of the bug you describe, btw. (on
> OS X
> intel or ppc). It does remind me of issues I had when attempting to
> use
> the GMP libs, though.
Was out all day; I'll look into it this weekend. Thanks for checking
up on it.
kid51
James Keenan
On Mar 9, 2007, at 10:42 AM, Will Coleda via RT wrote:
>
>
> Do you still get the bug when you revert the change to Configure.pl?
>
This morning, I checked out a fresh version from trunk: r17419. I
ran myconfigure.sh, which internally called the HEAD version of
Configure.pl.
#!/bin/sh
CC="/usr/bin/gcc-3.3"
CX="/usr/bin/g++-3.3"
/usr/local/bin/perl Configure.pl --cc="$CC" --cxx="$CX" --
link="$CX" --ld="$CX" --without-icu --without-gmp $@
The configuration built thereby is attached in 'myconfig.first.txt'.
I then ran 'make'. 'make' failed; here's the output starting at line
601:
make[1]: Leaving directory `/Users/jimk/work/fresh/docs'
src/nci_test.c
c++ -bundle -undefined suppress -L/usr/local/lib -L/Users/jimk/
work/fresh/blib/lib -flat_namespace \
-o runtime/parrot/dynext/libnci_test.bundle src/nci_test.o -lm
c++: couldn't run 'undle-gcc-4.0.3': No such file or directory
<-- line 605
make: *** [runtime/parrot/dynext/libnci_test.bundle] Error 1
Note the 'undle-gcc-4.0.3' in line 605. This misspelling is a
different instance of the same problem that we addressed with
'myconfigure.sh'.
I then reverted to the previous version of Configure.pl, which I
attach as 'my_Configure.pl'. I called 'make realclean',
'myconfigure.sh'. The resulting configuration created is attached as
'myconfig.second.txt', but here's the diff between the two myconfig's:
[fresh] 514 $ diff myconfig.first.txt myconfig.second.txt
2c2
< configdate='Sat Mar 10 09:40:26 2007'
---
> configdate='Sat Mar 10 10:05:47 2007'
12c12
< ld='c++', ldflags=' -L/usr/local/lib -L/Users/jimk/work/fresh/
blib/lib -flat_namespace ',
---
> ld='/usr/bin/g++-3.3', ldflags=' -L/usr/local/lib -L/Users/
jimk/work/fresh/blib/lib -flat_namespace ',
14c14
< libs=' -lm'
---
> libs='-lm'
I then re-ran 'make'. 'make' succeeded. All of the above -- problem
and (hack) solution -- is consistent with all my results over the
last two months since I originally reported the problem.
I should note that I can do a "straight-out-of-the-HEAD-box"
compilation on Linux with no additional configuration options,
hacking, etc.
HTH!
kid51
Will Coleda via RT
On Fri Mar 16 19:24:55 2007, jk...@verizon.net wrote:
> Results confirmed again at r17522 tonight.