Re: [PATCH][gdb/testsuite] Fix ada tests with -fPIE/-pie

[Tom de Vries <tdevries@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1620 bytes --]

On 07-08-19 17:27, Tom de Vries wrote:
> On 07-08-19 16:18, Tom Tromey wrote:
>>>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:
>>
>> Tom> When running the gdb testsuite with target board unix/-fPIE/-pie, the
>> Tom> resulting ada executables are not PIE executables, because gnatmake doesn't
>> Tom> recognize -pie, and consequently doesn't pass it to gnatlink.
>>
>> Tom> Fix this by replacing "-pie" with "-largs -pie -margs" for ada test-cases in
>> Tom> gdb_default_target_compile, and doing the same for -no-pie.
>>
>> I think this is a good idea overall.
>>
>> However, is gdb_default_target_compile still used?  And if so, by what
>> path?
> 
> I'm using dejagnu 1.6.1, and that one does not have find_go_linker. So,
> use_gdb_compile is set to 1 and we get:
> ...
> if {$use_gdb_compile} {
>     catch {rename default_target_compile {}}
>     rename gdb_default_target_compile default_target_compile
> }
> ...
> 
>>  My understanding is that with a "new enough" dejagnu, it won't be
>> used -- so some users might still see the old behavior.
>>
> 
> AFAIU, yes. Hmm, that's not good.
> 
>> Basically gdb_default_target_compile is all a big monkeypatching hack
> 
> /me reads https://en.wikipedia.org/wiki/Monkey_patch
> 
>> and it would be way better to have some kind of more principled approach
>> upstream.
> 
> Agreed.
> 
>> I don't know what that would look like.  And of course to get
>> there we'd probably need even more monkeypatching.
> 
> The following uses the approach taken in lib/cell.exp.
> 
> Is this any better?
> 

Updated rationale and ChangeLog entry.

OK for trunk?

Thanks,
- Tom


[-- Attachment #2: 0001-gdb-testsuite-Fix-ada-tests-with-fPIE-pie.patch --]
[-- Type: text/x-patch, Size: 1811 bytes --]

[gdb/testsuite] Fix ada tests with -fPIE/-pie

When running the gdb testsuite with target board unix/-fPIE/-pie, the
resulting ada executables are not PIE executables, because gnatmake doesn't
recognize -pie, and consequently doesn't pass it to gnatlink.

Fix this by replacing "-pie" with "-largs -pie -margs" in
target_compile_ada_from_dir, and doing the same for -no-pie.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2019-08-08  Tom de Vries  <tdevries@suse.de>

	PR testsuite/24888
	* lib/ada.exp (target_compile_ada_from_dir): Route -pie/-no-pie to
	gnatlink.

---
 gdb/testsuite/lib/ada.exp | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/gdb/testsuite/lib/ada.exp b/gdb/testsuite/lib/ada.exp
index 1345c747c5..6a3fd33240 100644
--- a/gdb/testsuite/lib/ada.exp
+++ b/gdb/testsuite/lib/ada.exp
@@ -19,11 +19,36 @@
 
 proc target_compile_ada_from_dir {builddir source dest type options} {
     set saved_cwd [pwd]
+
+    global board
+    set board [target_info name]
+    set save_multilib_flag [board_info $board multilib_flags]
+    set multilib_flag ""
+    foreach op $save_multilib_flag {
+	if { $op == "-pie" || $op == "-no-pie" } {
+	    # Pretend gnatmake supports -pie/-no-pie, route it to
+	    # linker.
+	    append multilib_flag " -largs $op -margs"
+	} else {
+	    append multilib_flag " $op"
+	}
+    }
+    if { $multilib_flag != "" } {
+	unset_board_info "multilib_flags"
+	set_board_info multilib_flags "$multilib_flag"
+    }
+
     catch {
         cd $builddir
         return [target_compile $source $dest $type $options]
     } result options
     cd $saved_cwd
+
+    if { $save_multilib_flag != "" } {
+	unset_board_info "multilib_flags"
+	set_board_info multilib_flags $save_multilib_flag
+    }
+
     return -options $options $result
 }