PATCH: Add check_lto_available

PATCH: Add check_lto_available

[H.J. Lu]

Hi,

I checked in this patch to add check_lto_available to check if
compiler supports LTO.


H.J.
---
Index: ChangeLog
===================================================================
RCS file: /cvs/src/src/ld/testsuite/ChangeLog,v
retrieving revision 1.1637
diff -u -p -r1.1637 ChangeLog
--- ChangeLog	7 Dec 2012 17:53:03 -0000	1.1637
+++ ChangeLog	7 Dec 2012 17:55:20 -0000
@@ -1,5 +1,10 @@
 2012-12-07  H.J. Lu  <hongjiu.lu@intel.com>
 
+	* lib/ld-lib.exp (check_lto_available): New.  Check if compiler
+	supports LTO.
+
+2012-12-07  H.J. Lu  <hongjiu.lu@intel.com>
+
 	* lib/ld-lib.exp (run_cc_link_tests): Properly check linker
 	warnings.
 
Index: lib/ld-lib.exp
===================================================================
RCS file: /cvs/src/src/ld/testsuite/lib/ld-lib.exp,v
retrieving revision 1.99
diff -u -p -r1.99 ld-lib.exp
--- lib/ld-lib.exp	7 Dec 2012 17:53:04 -0000	1.99
+++ lib/ld-lib.exp	7 Dec 2012 17:55:20 -0000
@@ -1614,6 +1614,34 @@ proc check_plugin_api_available { } {
     return $plugin_api_available_saved
 }
 
+# Returns true if the target compiler supports LTO
+proc check_lto_available { } {
+    global lto_available_saved
+    global CC
+    if {![info exists lto_available_saved]} {
+	# Check if gcc supports -flto -fuse-linker-plugin
+	if { [which $CC] == 0 } {
+	    set lto_available_saved 0
+	    return 0
+	}
+	set basename "lto"
+	set src ${basename}[pid].c
+	set output ${basename}[pid].so
+	set f [open $src "w"]
+	puts $f ""
+	close $f
+	set status [remote_exec host $CC "-shared -B[pwd]/tmpdir/ld/ -flto -fuse-linker-plugin $src -o $output"]
+	if { [lindex $status 0] == 0 } {
+	    set lto_available_saved 1
+	} else {
+	    set lto_available_saved 0
+	}
+	file delete $src
+	file delete $output
+    }
+    return $lto_available_saved
+}
+
 # Check if the assembler supports CFI statements.
 
 proc check_as_cfi { } {

Re: PATCH: Add check_lto_available

[Richard Earnshaw]

HJ,

Why does this patch try to produce a shared object as part of
determining whether LTO works?  If you have LTO but not shared object
support in a toolchain this means we don't test LTO properly.

R.

On 07/12/12 17:58, H.J. Lu wrote:
> Hi,
> 
> I checked in this patch to add check_lto_available to check if
> compiler supports LTO.
> 
> 
> H.J.
> ---
> Index: ChangeLog
> ===================================================================
> RCS file: /cvs/src/src/ld/testsuite/ChangeLog,v
> retrieving revision 1.1637
> diff -u -p -r1.1637 ChangeLog
> --- ChangeLog	7 Dec 2012 17:53:03 -0000	1.1637
> +++ ChangeLog	7 Dec 2012 17:55:20 -0000
> @@ -1,5 +1,10 @@
>  2012-12-07  H.J. Lu  <hongjiu.lu@intel.com>
>  
> +	* lib/ld-lib.exp (check_lto_available): New.  Check if compiler
> +	supports LTO.
> +
> +2012-12-07  H.J. Lu  <hongjiu.lu@intel.com>
> +
>  	* lib/ld-lib.exp (run_cc_link_tests): Properly check linker
>  	warnings.
>  
> Index: lib/ld-lib.exp
> ===================================================================
> RCS file: /cvs/src/src/ld/testsuite/lib/ld-lib.exp,v
> retrieving revision 1.99
> diff -u -p -r1.99 ld-lib.exp
> --- lib/ld-lib.exp	7 Dec 2012 17:53:04 -0000	1.99
> +++ lib/ld-lib.exp	7 Dec 2012 17:55:20 -0000
> @@ -1614,6 +1614,34 @@ proc check_plugin_api_available { } {
>      return $plugin_api_available_saved
>  }
>  
> +# Returns true if the target compiler supports LTO
> +proc check_lto_available { } {
> +    global lto_available_saved
> +    global CC
> +    if {![info exists lto_available_saved]} {
> +	# Check if gcc supports -flto -fuse-linker-plugin
> +	if { [which $CC] == 0 } {
> +	    set lto_available_saved 0
> +	    return 0
> +	}
> +	set basename "lto"
> +	set src ${basename}[pid].c
> +	set output ${basename}[pid].so
> +	set f [open $src "w"]
> +	puts $f ""
> +	close $f
> +	set status [remote_exec host $CC "-shared -B[pwd]/tmpdir/ld/ -flto -fuse-linker-plugin $src -o $output"]
> +	if { [lindex $status 0] == 0 } {
> +	    set lto_available_saved 1
> +	} else {
> +	    set lto_available_saved 0
> +	}
> +	file delete $src
> +	file delete $output
> +    }
> +    return $lto_available_saved
> +}
> +
>  # Check if the assembler supports CFI statements.
>  
>  proc check_as_cfi { } {
> 

Re: PATCH: Add check_lto_available

[H.J. Lu]

On Fri, Oct 11, 2013 at 5:36 AM, Richard Earnshaw <rearnsha@arm.com> wrote:
> HJ,
>
> Why does this patch try to produce a shared object as part of
> determining whether LTO works?  If you have LTO but not shared object
> support in a toolchain this means we don't test LTO properly.
>

LTO tests also check if shared library works correctly with LTO.

-- 
H.J.

Re: PATCH: Add check_lto_available

[Vidya Praveen]

On Fri, Oct 11, 2013 at 06:09:43PM +0100, H.J. Lu wrote:
> On Fri, Oct 11, 2013 at 5:36 AM, Richard Earnshaw <rearnsha@arm.com> wrote:
> > HJ,
> >
> > Why does this patch try to produce a shared object as part of
> > determining whether LTO works?  If you have LTO but not shared object
> > support in a toolchain this means we don't test LTO properly.
> >
> 
> LTO tests also check if shared library works correctly with LTO.

lto.exp does hav a condition to guard those:

303 if { [is_elf_format]
304      && [run_host_cmd_yesno $CC "-shared -fPIC $srcdir/$subdir/dummy.c -o tmpdir/t.so"] } {
305     run_cc_link_tests $lto_link_elf_tests
....
313 }

Since the other tests seem valid for toolchains that doesn't support shared
library, Shouldn't check_lto_availble have a check that doesn't use -shared ?

VP