Boris Ostrovsky
2012-Jun-08 20:23 UTC
[PATCH v2] tools, configure: Fix LIB_PATH computation in configure scripts
# HG changeset patch
# User Boris Ostrovsky <boris.ostrovsky@amd.com>
# Date 1339185838 -7200
# Node ID fa21f847fc66619fad38923cd87d6ba51d731eba
# Parent 32034d1914a607d7b6f1f060352b4cac973600f8
tools, configure: Fix LIB_PATH computation in configure scripts
tool''s configure script sets LIB_PATH by chopping off ${exec_prefix}
from $libdir and it does so by computing length of ${exec_prefix} value.
However, $libdir''s value is a literal
''${exec_prefix}/lib'' string
(i.e. $exec_prefix is not substituted) and therefore LIB_PATH may be
computed incorrectly, most likely as "c_prefix}/lib64" assuming that
exec_prefix is NONE.
Instead, we should start at offset `expr length
''${exec_prefix}/''
(which is 15).
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@amd.com>
diff -r 32034d1914a6 -r fa21f847fc66 tools/configure
--- a/tools/configure Thu Jun 07 19:46:57 2012 +0100
+++ b/tools/configure Fri Jun 08 22:03:58 2012 +0200
@@ -6062,7 +6062,7 @@ fi
else
- LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}"
+ LIB_PATH="${libdir:`expr length
''${exec_prefix}/''`}"
fi
diff -r 32034d1914a6 -r fa21f847fc66 tools/m4/default_lib.m4
--- a/tools/m4/default_lib.m4 Thu Jun 07 19:46:57 2012 +0100
+++ b/tools/m4/default_lib.m4 Fri Jun 08 22:03:58 2012 +0200
@@ -9,6 +9,6 @@ AC_DEFUN([AX_DEFAULT_LIB],
LIB_PATH="lib"
])
], [
- LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}"
+ LIB_PATH="${libdir:`expr length
''${exec_prefix}/''`}"
])
AC_SUBST(LIB_PATH)])
Ian Campbell
2012-Jul-04 11:00 UTC
Re: [PATCH v2] tools, configure: Fix LIB_PATH computation in configure scripts
On Fri, 2012-06-08 at 21:23 +0100, Boris Ostrovsky wrote:> # HG changeset patch > # User Boris Ostrovsky <boris.ostrovsky@amd.com> > # Date 1339185838 -7200 > # Node ID fa21f847fc66619fad38923cd87d6ba51d731eba > # Parent 32034d1914a607d7b6f1f060352b4cac973600f8 > tools, configure: Fix LIB_PATH computation in configure scripts > > tool''s configure script sets LIB_PATH by chopping off ${exec_prefix} > from $libdir and it does so by computing length of ${exec_prefix} value. > However, $libdir''s value is a literal ''${exec_prefix}/lib'' string > (i.e. $exec_prefix is not substituted) and therefore LIB_PATH may be > computed incorrectly, most likely as "c_prefix}/lib64" assuming that > exec_prefix is NONE. > > Instead, we should start at offset `expr length ''${exec_prefix}/'' > (which is 15). > > Signed-off-by: Boris Ostrovsky <boris.ostrovsky@amd.com>Boris, sorry for letting this languish for so long. Roger, any opinion on this patch? If I''m reading the script right this is the else clause of [AS_IF([test "\${exec_prefix}/lib" = "$libdir"], so at this point we know libdir != ${exec_prefix}/lib, is there any guarantee that it even starts with \${exec_prefix} ? I think this patch is complementary too rather than conflicting with / subsumed by Matt''s "tools: honour --libdir when it is passed to ./configure" ? Ian.> > diff -r 32034d1914a6 -r fa21f847fc66 tools/configure > --- a/tools/configure Thu Jun 07 19:46:57 2012 +0100 > +++ b/tools/configure Fri Jun 08 22:03:58 2012 +0200 > @@ -6062,7 +6062,7 @@ fi > > else > > - LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}" > + LIB_PATH="${libdir:`expr length ''${exec_prefix}/''`}" > > fi > > diff -r 32034d1914a6 -r fa21f847fc66 tools/m4/default_lib.m4 > --- a/tools/m4/default_lib.m4 Thu Jun 07 19:46:57 2012 +0100 > +++ b/tools/m4/default_lib.m4 Fri Jun 08 22:03:58 2012 +0200 > @@ -9,6 +9,6 @@ AC_DEFUN([AX_DEFAULT_LIB], > LIB_PATH="lib" > ]) > ], [ > - LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}" > + LIB_PATH="${libdir:`expr length ''${exec_prefix}/''`}" > ]) > AC_SUBST(LIB_PATH)]) > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
Roger Pau Monne
2012-Jul-04 11:26 UTC
Re: [PATCH v2] tools, configure: Fix LIB_PATH computation in configure scripts
Ian Campbell wrote:> On Fri, 2012-06-08 at 21:23 +0100, Boris Ostrovsky wrote: >> # HG changeset patch >> # User Boris Ostrovsky<boris.ostrovsky@amd.com> >> # Date 1339185838 -7200 >> # Node ID fa21f847fc66619fad38923cd87d6ba51d731eba >> # Parent 32034d1914a607d7b6f1f060352b4cac973600f8 >> tools, configure: Fix LIB_PATH computation in configure scripts >> >> tool''s configure script sets LIB_PATH by chopping off ${exec_prefix} >> from $libdir and it does so by computing length of ${exec_prefix} value. >> However, $libdir''s value is a literal ''${exec_prefix}/lib'' string >> (i.e. $exec_prefix is not substituted) and therefore LIB_PATH may be >> computed incorrectly, most likely as "c_prefix}/lib64" assuming that >> exec_prefix is NONE. >> >> Instead, we should start at offset `expr length ''${exec_prefix}/'' >> (which is 15). >> >> Signed-off-by: Boris Ostrovsky<boris.ostrovsky@amd.com> > > Boris, sorry for letting this languish for so long. > > Roger, any opinion on this patch?>> If I''m reading the script right this is the else clause of > [AS_IF([test "\${exec_prefix}/lib" = "$libdir"], > > so at this point we know libdir != ${exec_prefix}/lib, is there any > guarantee that it even starts with \${exec_prefix} ? > > I think this patch is complementary too rather than conflicting with / > subsumed by Matt''s "tools: honour --libdir when it is passed > to ./configure" ?I think this patch is not needed anymore if we apply Matt''s one, since Matt''s patch gets rid of LIB_PATH usage, and instead uses exec_prefix and libdir directly.> Ian. > >> diff -r 32034d1914a6 -r fa21f847fc66 tools/configure >> --- a/tools/configure Thu Jun 07 19:46:57 2012 +0100 >> +++ b/tools/configure Fri Jun 08 22:03:58 2012 +0200 >> @@ -6062,7 +6062,7 @@ fi >> >> else >> >> - LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}" >> + LIB_PATH="${libdir:`expr length ''${exec_prefix}/''`}" >> >> fi >> >> diff -r 32034d1914a6 -r fa21f847fc66 tools/m4/default_lib.m4 >> --- a/tools/m4/default_lib.m4 Thu Jun 07 19:46:57 2012 +0100 >> +++ b/tools/m4/default_lib.m4 Fri Jun 08 22:03:58 2012 +0200 >> @@ -9,6 +9,6 @@ AC_DEFUN([AX_DEFAULT_LIB], >> LIB_PATH="lib" >> ]) >> ], [ >> - LIB_PATH="${libdir:`expr length "$exec_prefix" + 1`}" >> + LIB_PATH="${libdir:`expr length ''${exec_prefix}/''`}" >> ]) >> AC_SUBST(LIB_PATH)]) >> >> >> _______________________________________________ >> Xen-devel mailing list >> Xen-devel@lists.xen.org >> http://lists.xen.org/xen-devel > > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
Ian Campbell
2012-Jul-04 11:27 UTC
Re: [PATCH v2] tools, configure: Fix LIB_PATH computation in configure scripts
On Wed, 2012-07-04 at 12:26 +0100, Roger Pau Monne wrote:> Ian Campbell wrote: > > On Fri, 2012-06-08 at 21:23 +0100, Boris Ostrovsky wrote: > >> # HG changeset patch > >> # User Boris Ostrovsky<boris.ostrovsky@amd.com> > >> # Date 1339185838 -7200 > >> # Node ID fa21f847fc66619fad38923cd87d6ba51d731eba > >> # Parent 32034d1914a607d7b6f1f060352b4cac973600f8 > >> tools, configure: Fix LIB_PATH computation in configure scripts > >> > >> tool''s configure script sets LIB_PATH by chopping off ${exec_prefix} > >> from $libdir and it does so by computing length of ${exec_prefix} value. > >> However, $libdir''s value is a literal ''${exec_prefix}/lib'' string > >> (i.e. $exec_prefix is not substituted) and therefore LIB_PATH may be > >> computed incorrectly, most likely as "c_prefix}/lib64" assuming that > >> exec_prefix is NONE. > >> > >> Instead, we should start at offset `expr length ''${exec_prefix}/'' > >> (which is 15). > >> > >> Signed-off-by: Boris Ostrovsky<boris.ostrovsky@amd.com> > > > > Boris, sorry for letting this languish for so long. > > > > Roger, any opinion on this patch? > > > > If I''m reading the script right this is the else clause of > > [AS_IF([test "\${exec_prefix}/lib" = "$libdir"], > > > > so at this point we know libdir != ${exec_prefix}/lib, is there any > > guarantee that it even starts with \${exec_prefix} ? > > > > I think this patch is complementary too rather than conflicting with / > > subsumed by Matt''s "tools: honour --libdir when it is passed > > to ./configure" ? > > I think this patch is not needed anymore if we apply Matt''s one, since > Matt''s patch gets rid of LIB_PATH usage, and instead uses exec_prefix > and libdir directly.So Matt''s patch could/should remove tools/m4/default_lib.m4? Ian.
Roger Pau Monne
2012-Jul-04 11:31 UTC
Re: [PATCH v2] tools, configure: Fix LIB_PATH computation in configure scripts
Ian Campbell wrote:> On Wed, 2012-07-04 at 12:26 +0100, Roger Pau Monne wrote: >> Ian Campbell wrote: >>> On Fri, 2012-06-08 at 21:23 +0100, Boris Ostrovsky wrote: >>>> # HG changeset patch >>>> # User Boris Ostrovsky<boris.ostrovsky@amd.com> >>>> # Date 1339185838 -7200 >>>> # Node ID fa21f847fc66619fad38923cd87d6ba51d731eba >>>> # Parent 32034d1914a607d7b6f1f060352b4cac973600f8 >>>> tools, configure: Fix LIB_PATH computation in configure scripts >>>> >>>> tool''s configure script sets LIB_PATH by chopping off ${exec_prefix} >>>> from $libdir and it does so by computing length of ${exec_prefix} value. >>>> However, $libdir''s value is a literal ''${exec_prefix}/lib'' string >>>> (i.e. $exec_prefix is not substituted) and therefore LIB_PATH may be >>>> computed incorrectly, most likely as "c_prefix}/lib64" assuming that >>>> exec_prefix is NONE. >>>> >>>> Instead, we should start at offset `expr length ''${exec_prefix}/'' >>>> (which is 15). >>>> >>>> Signed-off-by: Boris Ostrovsky<boris.ostrovsky@amd.com> >>> Boris, sorry for letting this languish for so long. >>> >>> Roger, any opinion on this patch? >> > >>> If I''m reading the script right this is the else clause of >>> [AS_IF([test "\${exec_prefix}/lib" = "$libdir"], >>> >>> so at this point we know libdir != ${exec_prefix}/lib, is there any >>> guarantee that it even starts with \${exec_prefix} ? >>> >>> I think this patch is complementary too rather than conflicting with / >>> subsumed by Matt''s "tools: honour --libdir when it is passed >>> to ./configure" ? >> I think this patch is not needed anymore if we apply Matt''s one, since >> Matt''s patch gets rid of LIB_PATH usage, and instead uses exec_prefix >> and libdir directly. > > So Matt''s patch could/should remove tools/m4/default_lib.m4?Yes, I''ve already replied to that thread also :)