[ patch, testsuite, fortran] Replace "call abort" by "stop n"

[ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Thomas Koenig]

[-- Attachment #1: Type: text/plain, Size: 573 bytes --]

Hello world,

the attached patch replaces the non-standard "call abort" by
"STOP n", where n is a number counted up from 1 for each occurrence.
This numbering was suggested by Neil Carlson.

This makes the gfortran testsuite _much_ more standard-conforming.
It also makes it easier to see _which_ of the numerous tests failed,
if anything happens.

I had to adjust the tree dump in a couple of places to check
for stop instead of abort.  In the standards tests where the
compiler was supposed to complain about call abort, I left it in :-)

OK for trunk?

Regards

	Thomas

[-- Attachment #2: p1.diff.gz --]
[-- Type: application/gzip, Size: 389995 bytes --]

[-- Attachment #3: ChangeLog.gz --]
[-- Type: application/gzip, Size: 10575 bytes --]

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Thomas Koenig]

[-- Attachment #1: Type: text/plain, Size: 194 bytes --]

I wrote:

> OK for trunk?

Following a remark by Steve, here is an additional test case which
checks if abort actually works.  I'll add that to the patch
when it is committed.

Regards

	Thomas

[-- Attachment #2: abort_shouldfail.f90 --]
[-- Type: text/x-fortran, Size: 99 bytes --]

! { dg-do  run }
! { dg-shouldfail "Program aborted." }
program main
  call abort
end program main

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Steve Kargl]

On Fri, Feb 16, 2018 at 12:25:11AM +0100, Thomas Koenig wrote:
> I wrote:
> 
> > OK for trunk?
> 
> Following a remark by Steve, here is an additional test case which
> checks if abort actually works.  I'll add that to the patch
> when it is committed.
> 

The patch is ok from my viewpoint.   You might want to
give others a chance to chime in.  

-- 
Steve

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Janus Weil]

2018-02-16 0:37 GMT+01:00 Steve Kargl <sgk@troutmask.apl.washington.edu>:
> On Fri, Feb 16, 2018 at 12:25:11AM +0100, Thomas Koenig wrote:
>> I wrote:
>>
>> > OK for trunk?
>>
>> Following a remark by Steve, here is an additional test case which
>> checks if abort actually works.  I'll add that to the patch
>> when it is committed.
>>
>
> The patch is ok from my viewpoint.   You might want to
> give others a chance to chime in.

Ok with me as well. (Personally I'd have gone with a lowercase "stop",
but uppercase is also fine.)

Thanks for taking care of this! The hardest part was probably the ChangeLog ;)

Cheers,
Janus

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Janne Blomqvist]

On Fri, Feb 16, 2018 at 12:30 AM, Thomas Koenig <tkoenig@netcologne.de> wrote:
> Hello world,
>
> the attached patch replaces the non-standard "call abort" by
> "STOP n", where n is a number counted up from 1 for each occurrence.
> This numbering was suggested by Neil Carlson.
>
> This makes the gfortran testsuite _much_ more standard-conforming.
> It also makes it easier to see _which_ of the numerous tests failed,
> if anything happens.
>
> I had to adjust the tree dump in a couple of places to check
> for stop instead of abort.  In the standards tests where the
> compiler was supposed to complain about call abort, I left it in :-)

Hi,

as a result of this, you should be able to remove the vast majority of
-fall-intrinsics extra options. Please do that as well.

I actually liked the usage of "call abort" in testcases, as a failing
testcase provided a backtrace with line number information and a core
dump. With this change, one now has to check on which line the
testcase failed, set a breakpoint there and run it again in the
debugger. Oh well, I guess that convenience has to be sacrificed on
the altar of standards conformance.

There's also the issue that without special tricks only the lower 8
bits of the exit code are used, and furthermore there are various
conventions. See e.g. https://stackoverflow.com/a/1535733/75652 for
what's typically the meaning on Linux. Anyway, I guess this is mostly
a theoretical quibble as I don't think we have any testcases with that
many error checks.


-- 
Janne Blomqvist

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Thomas Koenig]

Am 16.02.2018 um 09:11 schrieb Janne Blomqvist:
> as a result of this, you should be able to remove the vast majority of
> -fall-intrinsics extra options. Please do that as well.

Done, in r257782.

> I actually liked the usage of "call abort" in testcases, as a failing
> testcase provided a backtrace with line number information and a core
> dump. With this change, one now has to check on which line the
> testcase failed, set a breakpoint there and run it again in the
> debugger. Oh well, I guess that convenience has to be sacrificed on
> the altar of standards conformance.

Well, we get the output "STOP 1" for a STOP 1 statement.
Should not be too hard to find in the source.

Regards

	Thomas

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Janus Weil]

[-- Attachment #1: Type: text/plain, Size: 456 bytes --]

Hi Thomas,

>> as a result of this, you should be able to remove the vast majority of
>> -fall-intrinsics extra options. Please do that as well.
>
>
> Done, in r257782.

Regarding "-fall-intrinsics": Your commit has greatly reduced its
usage, but I still see a few cases that you left in, although the flag
does not really seem to be required.

Is there a reason why did not treat these? Would you mind if I applied
the patch in attachment?

Cheers,
Janus

[-- Attachment #2: fall_intrinsics.diff --]
[-- Type: text/plain, Size: 3273 bytes --]

Index: gcc/testsuite/gfortran.dg/io_real_boz2.f90
===================================================================
--- gcc/testsuite/gfortran.dg/io_real_boz2.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/io_real_boz2.f90	(working copy)
@@ -1,6 +1,6 @@
 ! { dg-do run }
 ! { dg-shouldfail "Real BOZ not allowed" }
-! { dg-options "-fall-intrinsics -std=f2003" }
+! { dg-options "-std=f2003" }
 ! Test for invalid (F95/F2003) writing of real with octal edit descriptor
 ! PR fortran/29625
 program real_boz
Index: gcc/testsuite/gfortran.dg/pointer_intent_3.f90
===================================================================
--- gcc/testsuite/gfortran.dg/pointer_intent_3.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/pointer_intent_3.f90	(working copy)
@@ -1,5 +1,5 @@
 ! { dg-do compile }
-! { dg-options "-std=f2003 -fall-intrinsics" }
+! { dg-options "-std=f2003" }
 ! { dg-shouldfail "Invalid code" }
 !
 ! Pointer intent test
Index: gcc/testsuite/gfortran.dg/proc_ptr_common_1.f90
===================================================================
--- gcc/testsuite/gfortran.dg/proc_ptr_common_1.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/proc_ptr_common_1.f90	(working copy)
@@ -1,5 +1,5 @@
 ! { dg-do run }
-! { dg-options "-std=f2003 -fall-intrinsics" }
+! { dg-options "-std=f2003" }
 !
 ! PR fortran/36592
 !
Index: gcc/testsuite/gfortran.dg/protected_3.f90
===================================================================
--- gcc/testsuite/gfortran.dg/protected_3.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/protected_3.f90	(working copy)
@@ -1,4 +1,4 @@
-! { dg-options "-std=f95 -fall-intrinsics" }
+! { dg-options "-std=f95" }
 ! PR fortran/23994
 !
 ! Test PROTECTED attribute. Within the module everything is allowed.
Index: gcc/testsuite/gfortran.dg/protected_4.f90
===================================================================
--- gcc/testsuite/gfortran.dg/protected_4.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/protected_4.f90	(working copy)
@@ -1,6 +1,6 @@
 ! { dg-do compile }
 ! { dg-shouldfail "Invalid Fortran 2003 code" }
-! { dg-options "-std=f2003 -fall-intrinsics" }
+! { dg-options "-std=f2003" }
 ! PR fortran/23994
 !
 ! Test PROTECTED attribute. Within the module everything is allowed.
Index: gcc/testsuite/gfortran.dg/protected_5.f90
===================================================================
--- gcc/testsuite/gfortran.dg/protected_5.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/protected_5.f90	(working copy)
@@ -1,6 +1,6 @@
 ! { dg-do compile }
 ! { dg-shouldfail "Invalid Fortran 2003 code" }
-! { dg-options "-std=f2003 -fall-intrinsics" }
+! { dg-options "-std=f2003" }
 ! PR fortran/23994
 !
 ! Test PROTECTED attribute. Within the module everything is allowed.
Index: gcc/testsuite/gfortran.dg/protected_6.f90
===================================================================
--- gcc/testsuite/gfortran.dg/protected_6.f90	(revision 257784)
+++ gcc/testsuite/gfortran.dg/protected_6.f90	(working copy)
@@ -1,6 +1,6 @@
 ! { dg-do compile }
 ! { dg-shouldfail "Invalid Fortran 2003 code" }
-! { dg-options "-std=f2003 -fall-intrinsics" }
+! { dg-options "-std=f2003" }
 ! PR fortran/23994
 !
 ! Test PROTECTED attribute. Within the module everything is allowed.

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Thomas Koenig]

Hi Janus,

> Regarding "-fall-intrinsics": Your commit has greatly reduced its
> usage, but I still see a few cases that you left in, although the flag
> does not really seem to be required.
> 
> Is there a reason why did not treat these?

I was specifically chaning those cases which contained STOP [1-9]
after my patch; I didn't look at any other use of -fall-intrinsics.

> Would you mind if I applied
> the patch in attachment?

Not at all - if the test cases pass without the option, just go ahead
and commit the patch.

Regards

	Thomas

Re: [ patch, testsuite, fortran] Replace "call abort" by "stop n"

[Janus Weil]

2018-02-18 1:38 GMT+01:00 Thomas Koenig <tkoenig@netcologne.de>:
> Hi Janus,
>
>> Regarding "-fall-intrinsics": Your commit has greatly reduced its
>> usage, but I still see a few cases that you left in, although the flag
>> does not really seem to be required.
>>
>> Is there a reason why did not treat these?
>
> I was specifically chaning those cases which contained STOP [1-9]
> after my patch; I didn't look at any other use of -fall-intrinsics.
>
>> Would you mind if I applied
>> the patch in attachment?
>
>
> Not at all - if the test cases pass without the option, just go ahead
> and commit the patch.

Alrighty, committed as r257789 (after making sure there are no failures).

Cheers,
Janus