public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: "Ulrich Weigand" <uweigand@de.ibm.com>
To: hjl.tools@gmail.com (H.J. Lu)
Cc: mark.kettenis@xs4all.nl (Mark Kettenis),
	jan.kratochvil@redhat.com,        Jon.Zhou@jdsu.com,
	gdb-patches@sourceware.org
Subject: [rfc] skip_ilp32_tests / skip_lp64_tests predicates (was: Re: support biarch gcore?)
Date: Fri, 09 Jul 2010 15:48:00 -0000	[thread overview]
Message-ID: <201007091548.o69FmNlk015436@d12av02.megacenter.de.ibm.com> (raw)
In-Reply-To: <AANLkTimQRQeRUPtRFiXuU7Z2RWfYItt07gOcZBeAYj9R@mail.gmail.com> from "H.J. Lu" at Jul 08, 2010 06:51:55 AM

H.J. Lu wrote:

> Very good point. There are some arch specific tests which are either
> 32bit or 64bit. Currently, we check target. It doesn't work -m32. We
> need something like
> 
> /* { dg-require-effective-target lp64 } */
> /* { dg-require-effective-target ilp32 } */

Right; I noticed this problem with one ppc64 test case as well.

I've tried to add testsuite predicates skip_ilp32_tests and skip_lp64_tests
modeled along the lines of the above GCC effective-target flags, and this
helps with this ppc64 test case.

Could you try whether you can use this to fix the i386/x86_64 tests?

I'd appreciate any comments on the approach below!

Bye,
Ulrich


ChangeLog:

	* lib/gdb.exp (skip_ilp32_tests): New procedure.
	(skip_lp64_tests): New procedure.
	* gdb.arch/ppc64-atomic-inst.exp: Use skip_lp64_tests to execute test
	only when building 64-bit executables.  Do not hard-code -m64 option.

Index: gdb/testsuite/gdb.arch/ppc64-atomic-inst.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.arch/ppc64-atomic-inst.exp,v
retrieving revision 1.5
diff -u -p -r1.5 ppc64-atomic-inst.exp
--- gdb/testsuite/gdb.arch/ppc64-atomic-inst.exp	5 May 2010 18:06:57 -0000	1.5
+++ gdb/testsuite/gdb.arch/ppc64-atomic-inst.exp	9 Jul 2010 15:40:41 -0000
@@ -25,7 +25,7 @@ if $tracelevel {
 # instruction.
 
 
-if ![istarget "powerpc64-*"] then {
+if {![istarget "powerpc*"] || [skip_lp64_tests]} {
     verbose "Skipping testing of powerpc64 single stepping over atomic sequences."
     return
 }
@@ -33,9 +33,9 @@ if ![istarget "powerpc64-*"] then {
 set testfile "ppc64-atomic-inst"
 set srcfile ${testfile}.c
 set binfile ${objdir}/${subdir}/${testfile}
-set compile_flags "debug quiet"
+set compile_flags {debug quiet}
 
-if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable [concat $compile_flags additional_flags=-m64]] != "" } {
+if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable $compile_flags] != "" } {
     unsupported "Testcase compile failed."
     return -1
 }
Index: gdb/testsuite/lib/gdb.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/lib/gdb.exp,v
retrieving revision 1.152
diff -u -p -r1.152 gdb.exp
--- gdb/testsuite/lib/gdb.exp	28 Jun 2010 20:09:59 -0000	1.152
+++ gdb/testsuite/lib/gdb.exp	9 Jul 2010 15:40:42 -0000
@@ -1401,6 +1401,83 @@ proc skip_shlib_tests {} {
     return 1
 }
 
+# Return 1 if we should skip tests that assume the ILP32 model.
+# This cannot be decided simply from looking at the target string,
+# as it might depend on externally passed compiler options like -m64.
+proc skip_ilp32_tests {} {
+    global skip_ilp32_tests_saved
+
+    # Use the cached value, if it exists.  Cache value per "board" to handle
+    # runs with multiple options (e.g. unix/{-m32,-64}) correctly.
+    set me "skip_ilp32_tests"
+    set board [target_info name]
+    if [info exists skip_ilp32_tests_saved($board)] {
+        verbose "$me:  returning saved $skip_ilp32_tests_saved($board)" 2
+        return $skip_ilp32_tests_saved($board)
+    }
+
+
+    set src ilp32[pid].c
+    set obj ilp32[pid].o
+
+    set f [open $src "w"]
+    puts $f "int dummy\[sizeof (int) == 4"
+    puts $f "           && sizeof (void *) == 4"
+    puts $f "           && sizeof (long) == 4 ? 1 : -1\];"
+    close $f
+
+    verbose "$me:  compiling testfile $src" 2
+    set lines [gdb_compile $src $obj object {quiet}]
+    file delete $src
+    file delete $obj
+
+    if ![string match "" $lines] then {
+        verbose "$me:  testfile compilation failed, returning 1" 2
+        return [set skip_ilp32_tests_saved($board) 1]
+    }
+
+    verbose "$me:  returning 0" 2
+    return [set skip_ilp32_tests_saved($board) 0]
+}
+
+# Return 1 if we should skip tests that assume the LP64 model.
+# This cannot be decided simply from looking at the target string,
+# as it might depend on externally passed compiler options like -m64.
+proc skip_lp64_tests {} {
+    global skip_lp64_tests_saved
+
+    # Use the cached value, if it exists.  Cache value per "board" to handle
+    # runs with multiple options (e.g. unix/{-m32,-64}) correctly.
+    set me "skip_lp64_tests"
+    set board [target_info name]
+    if [info exists skip_lp64_tests_saved($board)] {
+        verbose "$me:  returning saved $skip_lp64_tests_saved($board)" 2
+        return $skip_lp64_tests_saved($board)
+    }
+
+    set src lp64[pid].c
+    set obj lp64[pid].o
+
+    set f [open $src "w"]
+    puts $f "int dummy\[sizeof (int) == 4"
+    puts $f "           && sizeof (void *) == 8"
+    puts $f "           && sizeof (long) == 8 ? 1 : -1\];"
+    close $f
+
+    verbose "$me:  compiling testfile $src" 2
+    set lines [gdb_compile $src $obj object {quiet}]
+    file delete $src
+    file delete $obj
+
+    if ![string match "" $lines] then {
+        verbose "$me:  testfile compilation failed, returning 1" 2
+        return [set skip_lp64_tests_saved($board) 1]
+    }
+
+    verbose "$me:  returning 0" 2
+    return [set skip_lp64_tests_saved($board) 0]
+}
+
 # Run a test on the target to see if it supports vmx hardware.  Return 0 if so, 
 # 1 if it does not.  Based on 'check_vmx_hw_available' from the GCC testsuite.
 


-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com

  reply	other threads:[~2010-07-09 15:48 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <201007062128.o66LSkNC032580@glazunov.sibelius.xs4all.nl>
2010-07-07 12:30 ` support biarch gcore? Ulrich Weigand
2010-07-08  2:35   ` Jon Zhou
2010-07-08 11:17     ` Ulrich Weigand
2010-07-08  4:47   ` H.J. Lu
2010-07-08  5:05     ` H.J. Lu
2010-07-08 11:15       ` Ulrich Weigand
2010-07-08 13:52         ` H.J. Lu
2010-07-09 15:48           ` Ulrich Weigand [this message]
2010-07-09 15:56             ` [rfc] skip_ilp32_tests / skip_lp64_tests predicates (was: Re: support biarch gcore?) H.J. Lu
2010-07-09 21:43             ` H.J. Lu
2010-07-09 22:33               ` H.J. Lu
2010-07-12 14:49                 ` Ulrich Weigand
2010-07-21 22:45           ` support biarch gcore? Joseph S. Myers

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=201007091548.o69FmNlk015436@d12av02.megacenter.de.ibm.com \
    --to=uweigand@de.ibm.com \
    --cc=Jon.Zhou@jdsu.com \
    --cc=gdb-patches@sourceware.org \
    --cc=hjl.tools@gmail.com \
    --cc=jan.kratochvil@redhat.com \
    --cc=mark.kettenis@xs4all.nl \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).