Mail #13615

Back to main page of archive

Previous mail   Next mail   Unformatted/full headers
Overview  10 days   Subject   Date   Thread   Author  

From: Gert Doering
Subject: New (gpc-20060325)
Date: 26 May 2006, 14:14:05

Hi Frank,

thanks for your quick reply - and sorry for being late at my end, I got
distracted by non-computer things.

On Fri, May 19, 2006 at 03:45:06PM +0200, Frank Heckenbach wrote:
> > I just tried it on AIX 5.3, with gcc-3.4.6, and the result is sort of
> > "mixed":
> > 
> > There is one line that needs to be changed in gcc/p/test/test_run,
> > because otherwise I'll get an error from the shell (in make check):
> > 
> > old:
> > 
> > echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo "(using GP)"; fi`"
> > 
> > new: 
> > 
> > echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo \"(using GP)\"; fi`"
> 
> What's that? The original line is:
> 
> echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo "(using GP)"; fi`"
> 
> Perhaps you unintentionally duplicated a part of it (from the
> beginning to the first "[")?

Yes.  I'm not sure how that happened, but that's what you get when using
the mouse for "insert lines of code into mail".

The (changed) line indeed looks like this:

echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo \"(using GP)\"; fi`"

> So I assume it should be changed to:
> 
> echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo \"(using GP)\"; fi`"

Yes.

> > (the change is that "(using GP)" needs extra quotes, otherwise the shell
> > will complain - this is with AIX' ksh, but as far as I know, recent bash
> > versions will be similarily picky)
> 
> Not AFAICS:
> 
> # bash --version
> GNU bash, version 3.1.0(1)-release (i686-pc-linux-gnu)
> Copyright (C) 2005 Free Software Foundation, Inc.
> # echo "`echo "foo"`"
> foo

Interesting.

Even more interesting is that the abovementioned "echo" command runs
under AIX's ksh and /bin/sh just fine - but that's to be expected, the
quotes just disappear, and then all you have left is 

 echo foo

which works :)

Things look different if you do this:

gd@hilb31:/s1/gcc-3.4.6/gcc/p/test> echo "`echo "(foo)"`"
ksh: 0403-057 Syntax error: `(' is not expected.
gd@hilb31:/s1/gcc-3.4.6/gcc/p/test> echo "`echo \"(foo)\"`"
(foo)

- but you're right, bash doesn't care one way or the other.


> I think it's not so much about being picky, but supporting a
> slightly extended syntax ("" within `` pairs, while ksh apparently
> considers the first unquoted " to be the closing one without regard
> to nesting). But anyway, the extra \'s don't hurt bash (and I hope
> no other shell either).

Unfortunately it's not so easy.  NetBSD's ksh does it just the other
way:

gert@kirk:/rhome/gert$ echo "`echo \"(foo)\"`"
/bin/ksh: syntax error: `(' unexpected
gert@kirk:/rhome/gert$ echo "`echo "(foo)"`"  
(foo)

*argh*

so maybe portable code really should avoid using the same quote character
on different levels.  Which means, the "right" thing to do would be:

echo "$PC `$PC $PFLAGS -dumpversion`, flags: $PFLAGS_NO_PATHS `if [ x"$GP" != x ]; then echo '(using GP)'; fi`"

- which seems to be fully portable to all shells I have tried.


> > After that, the test suite starts, but *all* test programs fail:
> > 
> > "gmake pascal.check-long" explains what is going wrong:
> > 
> > Test Run By gd on 2006-05-19 13:43:03
> > Native configuration is powerpc-ibm-aix5.3.0.0 (hilb31)
> > /s1/gpc-build-20060325-3.4.6/gcc/xgpc -B/s1/gpc-build-20060325-3.4.6/gcc/ 20060325, based on gcc-3.4.6, flags: -g -O3 -W -Wall -Wno-unused   
> > GPC-TEST-BEGIN
> > ==========================
> > TEST    abso1.pas:      cc1: warning: command line option "-funit-path=/gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/units" is valid for Pascal but not for C
> > cc1: warning: command line option "-fno-unit-path" is valid for Pascal but not for C
> > cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/units" is valid for Pascal but not for C
> > cc1: warning: command line option "-fautobuild" is valid for Pascal but not for C
> > cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/test" is valid for Pascal but not for C
> > cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/test/../rts" is valid for Pascal but not for C
> > cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/test/../units" is valid for Pascal but not for C
> > cc1: warning: command line option "-fexecutable-path=." is valid for Pascal but not for C
> > OK
> >
> > I'm not sure how to proceed here.  Why is it calling cc1?  Why is it passing
> > options that the backend doesn't like?  Is this something that gcc-3.4.6
> > has changed?
> 
> For debugging, you could insert an echo statement in test_run, to
> find out how exactly it invokes GPC, then try this command-line
> manually, and if it still fails, play with the options to find out
> what's causing this.
> 
> The relevant line for most test programs (including abso1) is:
> 
>     if { $PC_WITH_FLAGS -Werror "$1"; } 2>&1; then
> 
> Note there are some similar looking lines, this is line 254. So you
> could put before it:
> 
>     echo $PC_WITH_FLAGS -Werror "$1"

OK, here we go.  This is how it's called:

---PC_WITH_FLAGS---
/s1/gpc-build-20060325-3.4.6/gcc/xgpc -B/s1/gpc-build-20060325-3.4.6/gcc/ --unit-path=/gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/units -I /gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/units -I ../rts --no-unit-path --unit-path=/s1/gcc-3.4.6/gcc/p/units --autobuild   -g -O3 -W -Wall -Wno-unused   -o a.out     --unit-path=/s1/gcc-3.4.6/gcc/p/test --unit-path=/s1/gcc-3.4.6/gcc/p/test/../rts     --unit-path=/s1/gcc-3.4.6/gcc/p/test/../units -I /s1/gcc-3.4.6/gcc/p/test -I /s1/gcc-3.4.6/gcc/p/test/../units --executable-path=.      -Werror /s1/gcc-3.4.6/gcc/p/test/abso1.pas
---PC_WITH_FLAGS---

If I run that on "hello.pas" I get the same warnings from "cc1".

If I remove all "--unit-path" directives, and the "--executable-path" and 
"--no-unit-path" directives, the warnings are gone (which was to be 
expected, but isn't very enlightening).


Adding "-v" to the gpc command line suggests that the warnings might come
from "collect2":

...
GNU Pascal Compiler PreProcessor version 20060325, based on gcc-3.4.6

{$include "..."} search starts here:
{$include <...>} search starts here:
 /gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/units
 ../rts
 /s1/gcc-3.4.6/gcc/p/test
 /s1/gcc-3.4.6/gcc/p/test/../units
End of search list.
 as -u -mppc -o /tmp//cchPIwsI.o /tmp//cckcaRBw.s
 /s1/gpc-build-20060325-3.4.6/gcc/collect2 -bpT:0x10000000 -bpD:0x20000000 -btextro -bnodelcsect -bexport:/usr/lib/libg.exp -o a.out /lib/crt0.o -L/s1/gpc-build-20060325-3.4.6/gcc -L/gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6 -L/gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/../../.. /tmp//cchPIwsI.o -lgpc -lm /s1/gpc-build-20060325-3.4.6/gcc/libgcc.a /s1/gpc-build-20060325-3.4.6/gcc/libgcc_eh.a -lg -lc /s1/gpc-build-20060325-3.4.6/gcc/libgcc.a /s1/gpc-build-20060325-3.4.6/gcc/libgcc_eh.a
cc1: warning: command line option "-funit-path=/gnulocal/lib/gcc/powerpc-ibm-aix5.3.0.0/3.4.6/units" is valid for Pascal but not for C
cc1: warning: command line option "-fno-unit-path" is valid for Pascal but not for C
cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/units" is valid for Pascal but not for C
cc1: warning: command line option "-fautobuild" is valid for Pascal but not for C
cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/test" is valid for Pascal but not for C
cc1: warning: command line option "-funit-path=/s1/gcc-3.4.6/gcc/p/test/../rts" is valid for Pascal but not for C


Which is backed by the fact that if I run GPC als "compile-only" (gpc -c),
I don't get any warnings.


I still don't understand it, though.  Where does collect2 get these options
from?  Are they included in the .o file?

> > Compiling a simple "hello,world" program works, with no warnings:
> > 
> > gd@hilb31:/tmp> gpc -Wall -o hello hello.pas   
> > gd@hilb31:/tmp> ./hello                        
> > Hello, World
> 
> BTW, you made sure this is the newly installed GPC here, not an
> older version, did you?

Yes, there was no gpc on that machine beforehand.

It has the same problems, though, when I add some of the triggering options:

gd@hilb31:/tmp> /gnulocal/bin/gpc -Wall -o hello hello.pas               
gd@hilb31:/tmp> ./hello
Hello, World

gd@hilb31:/tmp> /gnulocal/bin/gpc -Wall --no-unit-path -o hello hello.pas
cc1: warning: command line option "-fno-unit-path" is valid for Pascal but not for C
gd@hilb31:/tmp> ./hello
Hello, World


So - what can I try next?

regards,

gert

Previous mail   Next mail   Unformatted/full headers
Overview  10 days   Subject   Date   Thread   Author  

Replies

In reply to

Back to main page of archive

Note: This page contains information that does not originate from the owner of this web site, but from the authors of the mails archived. The owner of this web site is not responsible for the content of such information. Any use of that infomation requires the consent of the respective author.

Where WWW addresses (URLs) in the mails archived are marked as hyperlinks, this is only for the comfort of the reader. The content of the web pages linked to like this does not necessarily reflect the opinion of the owner of this web site or of the authors of the mails archived. The owner of this web site is not responsible for the content of such web pages. Those pages are explicitly not to be considered as part of the content of this page, but merely as references.

This page was created by Crystal 0.999 (Linux 2.4.27/i686).