[obvious patch] add missing chmod to when copying headers

[obvious patch] add missing chmod to when copying headers

[Chris Demetriou]

[-- Attachment #1: Type: text/plain, Size: 672 bytes --]

odd corner case when making/installing: if sources were checked out
with umask that denies group or other read (e.g. 077 or 027), most of
the installed gcc includes would come out group/other-readable but a
few wouldn't.
problem: missing chmods after copying the headers, in stmp-int-headers.
Since all of the rest of the headers are chmodded after copy, the
solution is obvious.


bootstrapped, verified permissions manually, committed as obvious.


chris
---
[gcc/ChangeLog]
2010-03-11  Chris Demetriou  <cgd@google.com>

        * Makefile.in (stmp-int-hdrs): Make include/unwind.h,
        include/stdint-gcc.h, and include/stdint.h world-readable.

[-- Attachment #2: chmod.patch --]
[-- Type: text/x-diff, Size: 1027 bytes --]

[gcc/ChangeLog]
2010-03-11  Chris Demetriou  <cgd@google.com>

	* Makefile.in (stmp-int-hdrs): Make include/unwind.h,
	include/stdint-gcc.h, and include/stdint.h world-readable.

Index: gcc/Makefile.in
===================================================================
--- gcc/Makefile.in	(revision 157391)
+++ gcc/Makefile.in	(working copy)
@@ -3835,14 +3835,17 @@
 	done
 	rm -f include/unwind.h
 	cp $(UNWIND_H) include/unwind.h
+	chmod a+r include/unwind.h
 	rm -f include/stdint.h
 	if [ $(USE_GCC_STDINT) = wrap ]; then \
 	  rm -f include/stdint-gcc.h; \
 	  cp $(srcdir)/ginclude/stdint-gcc.h include/stdint-gcc.h; \
+	  chmod a+r include/stdint-gcc.h; \
 	  cp $(srcdir)/ginclude/stdint-wrap.h include/stdint.h; \
 	elif [ $(USE_GCC_STDINT) = provide ]; then \
 	  cp $(srcdir)/ginclude/stdint-gcc.h include/stdint.h; \
 	fi
+	chmod a+r include/stdint.h
 	set -e; for ml in `cat fixinc_list`; do \
 	  sysroot_headers_suffix=`echo $${ml} | sed -e 's/;.*$$//'`; \
 	  multi_dir=`echo $${ml} | sed -e 's/^[^;]*;//'`; \

Re: [obvious patch] add missing chmod to when copying headers

[Basile Starynkevitch]

Chris Demetriou wrote:
> odd corner case when making/installing: if sources were checked out
> with umask that denies group or other read (e.g. 077 or 027), most of
> the installed gcc includes would come out group/other-readable but a
> few wouldn't.


Perhaps the same should be done for PLUGIN_HEADERS (if not done yet).

Cheers

-- 
Basile STARYNKEVITCH         http://starynkevitch.net/Basile/
email: basile<at>starynkevitch<dot>net mobile: +33 6 8501 2359
8, rue de la Faiencerie, 92340 Bourg La Reine, France
*** opinions {are only mines, sont seulement les miennes} ***

Re: [obvious patch] add missing chmod to when copying headers

[Chris Demetriou]

On Thu, Mar 11, 2010 at 13:28, Basile Starynkevitch
<basile@starynkevitch.net> wrote:
>
> Chris Demetriou wrote:
>>
>> odd corner case when making/installing: if sources were checked out
>> with umask that denies group or other read (e.g. 077 or 027), most of
>> the installed gcc includes would come out group/other-readable but a
>> few wouldn't.
>
> Perhaps the same should be done for PLUGIN_HEADERS (if not done yet).

AFAICT they are installed using $(install_data), so get reasonable perms, e.g.:

/usr/bin/install -c -m 644 ../../trunk/gcc/coretypes.h
/g/users/cgd/proj/gcc-trunk/bld/../inst/lib/gcc/x86_64-linux/4.5.0/plugin/include/coretypes.h


FYI, IIRC there are a few files that get built/installed with the
umask of the person doing the build.  This is probably is a bug in the
build/install process (IMO at least), but is easily fixable by using a
'good' umask when building.  (the files whose permissions I fixed,
OTOH, came from umask at *checkout* time... somewhat confusing, and
the way the makefile installed the rest of the related headers was
obviously different so IMO there was an obvious fix to make.  I've not
looked into other issues, my *build* scripts set a good umask.  8-)



chris

Re: [obvious patch] add missing chmod to when copying headers

[Paolo Bonzini]

On 03/11/2010 09:31 PM, Chris Demetriou wrote:
> odd corner case when making/installing: if sources were checked out
> with umask that denies group or other read (e.g. 077 or 027), most of
> the installed gcc includes would come out group/other-readable but a
> few wouldn't.
> problem: missing chmods after copying the headers, in stmp-int-headers.
> Since all of the rest of the headers are chmodded after copy, the
> solution is obvious.
>
>
> bootstrapped, verified permissions manually, committed as obvious.

... but why didn't you use $(INSTALL_DATA)? :-)

Paolo

Re: [obvious patch] add missing chmod to when copying headers

[Chris Demetriou]

On Fri, Mar 12, 2010 at 02:23, Paolo Bonzini <bonzini@gnu.org> wrote:
>> bootstrapped, verified permissions manually, committed as obvious.
>
> ... but why didn't you use $(INSTALL_DATA)? :-)

*chuckle*  there are actually reasons, believe it or not!  Just
because it's obvious doesn't mean I didn't think about it.  8-)

(a) because the rest of the code here doesn't, and

(b) this *isn't* actually *installing* anything!  it's populating an
include directory internal to the build process.  The actual header
installation happens thanks to the install-headers-* targets AFAICT
(which don't use INSTALL_DATA either 8-).  It's not obvious to me
(esp. in light of the way the rest of the code works) that
INSTALL_DATA is appropriate here.


Changing install-headers-* may be appropriate... but it seems obvious
that a bunch of work was put into its current implementation as-is
(three separate implementations!).  At *minimum* that would be
non-obvious, whereas making all the headers copied in stmp-int-hdrs be
chmodded consistently *is* obvious.  8-)



chris

Re: [obvious patch] add missing chmod to when copying headers

[Paolo Bonzini]

On Fri, Mar 12, 2010 at 19:51, Chris Demetriou <cgd@google.com> wrote:
> On Fri, Mar 12, 2010 at 02:23, Paolo Bonzini <bonzini@gnu.org> wrote:
>>> bootstrapped, verified permissions manually, committed as obvious.
>>
>> ... but why didn't you use $(INSTALL_DATA)? :-)
>
> *chuckle*  there are actually reasons, believe it or not!  Just
> because it's obvious doesn't mean I didn't think about it

Thanks for answering.

Paolo

Re: [obvious patch] add missing chmod to when copying headers

[Eric Botcazou]

> 2010-03-11  Chris Demetriou  <cgd@google.com>
>
>         * Makefile.in (stmp-int-hdrs): Make include/unwind.h,
>         include/stdint-gcc.h, and include/stdint.h world-readable.

This breaks platforms without include/stdint.h though:
  http://gcc.gnu.org/ml/gcc/2009-07/msg00625.html

if [ none = wrap ]; then \
           rm -f include/stdint-gcc.h; \
           cp ../../gcc-head-src/gcc/ginclude/stdint-gcc.h 
include/stdint-gcc.h;
\
           chmod a+r include/stdint-gcc.h; \
           cp ../../gcc-head-src/gcc/ginclude/stdint-wrap.h 
include/stdint.h; \
         elif [ none = provide ]; then \
           cp ../../gcc-head-src/gcc/ginclude/stdint-gcc.h 
include/stdint.h; \
         fi
chmod a+r include/stdint.h
chmod: cannot access `include/stdint.h': No such file or directory
make[2]: *** [stmp-int-hdrs] Error 1

-- 
Eric Botcazou

Re: [obvious patch] add missing chmod to when copying headers

[Chris Demetriou]

On Mon, Mar 15, 2010 at 00:45, Eric Botcazou <ebotcazou@adacore.com> wrote:
> This breaks platforms without include/stdint.h though:
>  http://gcc.gnu.org/ml/gcc/2009-07/msg00625.html

Ah, crud.  I misread the 'if / 'elif' as 'if' / else'.

I'll check in a fix shortly.

Sorry about that.


chris

Re: [obvious patch] add missing chmod to when copying headers

[Chris Demetriou]

On Mon, Mar 15, 2010 at 00:49, Chris Demetriou <cgd@google.com> wrote:
> On Mon, Mar 15, 2010 at 00:45, Eric Botcazou <ebotcazou@adacore.com> wrote:
>> This breaks platforms without include/stdint.h though:
>>  http://gcc.gnu.org/ml/gcc/2009-07/msg00625.html
>
> Ah, crud.  I misread the 'if / 'elif' as 'if' / else'.
>
> I'll check in a fix shortly.

Fix committed as revision 157454.
(http://gcc.gnu.org/ml/gcc-patches/2010-03/msg00565.html)


thanks,

chris

Re: [obvious patch] add missing chmod to when copying headers

[Eric Botcazou]

> Fix committed as revision 157454.
> (http://gcc.gnu.org/ml/gcc-patches/2010-03/msg00565.html)

Thanks a lot for the quick turn around!

-- 
Eric Botcazou