[Bug testsuite/21910] New: Files from previous runs can make some testcases fail

[Bug testsuite/21910] New: Files from previous runs can make some testcases fail

[e9925248 at stud4 dot tuwien dot ac dot at]

If old dump files for a testcase are left, if the name of the dump file has
changed, errors like the following occur:

ERROR: gcc.dg/tree-ssa/structopt-1.c: error executing dg-final: couldn't open
"structopt-1.c.t50.lim structopt-1.c.t53.lim": no such file or directory
UNRESOLVED: gcc.dg/tree-ssa/structopt-1.c: error executing dg-final: couldn't
open "structopt-1.c.t50.lim structopt-1.c.t53.lim": no such file or directory

Deleting all generated *.c.t* files in the testsuite directory, solves this problem.

I use dejagnu from the src-tree of sourceware.org.

-- 
           Summary: Files from previous runs can make some testcases fail
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: testsuite
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: e9925248 at stud4 dot tuwien dot ac dot at
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910

[Bug testsuite/21910] Files from previous runs can make some testcases fail

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-04 15:11 -------
We have "/* { dg-final { cleanup-tree-dump "lim" } } */" in the testcase so it should have cleaned up, 
when did you get the version of HEAD?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910

[Bug testsuite/21910] Files from previous runs can make some testcases fail

[e9925248 at stud4 dot tuwien dot ac dot at]

------- Additional Comments From e9925248 at stud4 dot tuwien dot ac dot at  2005-06-05 13:01 -------
The example was taken out of a test log created at the beginning of May. The
current log where simliar failures for a lot files occured was not available any
more, because I had overwritten it with a new run, after deleting the files
causing the errors to verify the solution.

To problem is, that the testsuite searches for pattern XXXX.t??.XXX, if it wants
to open a dump file, and try to open the result. If multiple files match this
pattern, such an error occurs (see lib/scantree.exp).

To reproduce it on a clean directory, create a file named ssa-sink-1.c.t46.sink
(such a file was created some time ago by a testsuite run, because some error
occured while checking this testcase) in the testsuite run directory (where
gcc.sum is put). A make check prints then the following error:
ERROR: gcc.dg/tree-ssa/ssa-sink-1.c: error executing dg-final: couldn't open
"ssa-sink-1.c.t46.sink ssa-sink-1.c.t52.sink": no such file or directory

It leaves additionally ssa-sink-1.c.t52.sink in the directory. If the number of
the sink pass would change again, a third file would appear in the error message.

I know, this is not a severe error, as a rm -f testsuite/*.c.t* before each make
check solves the problem.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910

[Bug testsuite/21910] Files from previous runs can make some testcases fail

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-05 15:28 -------
Now what I am saying this is all fixed already with cleanup-tree-dump.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910

[Bug testsuite/21910] Files from previous runs can make some testcases fail

[e9925248 at stud4 dot tuwien dot ac dot at]

------- Additional Comments From e9925248 at stud4 dot tuwien dot ac dot at  2005-06-06 07:51 -------
In the normal case (no error happens), cleanup-tree-dump will delete such files,
as a more verbose run of the testsuite shows:
Testing tree-ssa/ssa-sink-2.c
doing compile
Invoking the compiler as /tmp/gcc/host-i686-pc-linux-gnu/gcc/xgcc
-B/tmp/gcc/host-i686-pc-linux-gnu/gcc/
/tmp/gcc/gcc/testsuite/gcc.dg/tree-ssa/ssa-sink-2.c   -O2 -fdump-tree-sink-stats
-fno-show-column -S  -o ssa-sink-2.s
Executing on host: /tmp/gcc/host-i686-pc-linux-gnu/gcc/xgcc
-B/tmp/gcc/host-i686-pc-linux-gnu/gcc/
/tmp/gcc/gcc/testsuite/gcc.dg/tree-ssa/ssa-sink-2.c   -O2 -fdump-tree-sink-stats
-fno-show-column -S  -o ssa-sink-2.s    (timeout = 300)
pid is 14380 -14380
waitres is 14380 exp6 0 0
output is
Checking pattern "sparc-*-sunos*" with i686-pc-linux-gnu
Checking pattern "alpha*-*-*" with i686-pc-linux-gnu
Checking pattern "hppa*-*-hpux*" with i686-pc-linux-gnu
Checking pattern "sparc-*-sunos*" with i686-pc-linux-gnu
Checking pattern "alpha*-*-*" with i686-pc-linux-gnu
Checking pattern "hppa*-*-hpux*" with i686-pc-linux-gnu
remove-build-file `ssa-sink-2.c.t??.sink'
remove-build-file `ssa-sink-2.c.t52.sink'

In my case, cleanup-tree-dump is not executed, as scan-tree-dump-times fails
before, because of the additonal file:
Testing tree-ssa/ssa-sink-1.c
doing compile
Invoking the compiler as /tmp/gcc/host-i686-pc-linux-gnu/gcc/xgcc
-B/tmp/gcc/host-i686-pc-linux-gnu/gcc/
/tmp/gcc/gcc/testsuite/gcc.dg/tree-ssa/ssa-sink-1.c   -O2 -fdump-tree-sink-stats
-fno-show-column -S  -o ssa-sink-1.s
Executing on host: /tmp/gcc/host-i686-pc-linux-gnu/gcc/xgcc
-B/tmp/gcc/host-i686-pc-linux-gnu/gcc/
/tmp/gcc/gcc/testsuite/gcc.dg/tree-ssa/ssa-sink-1.c   -O2 -fdump-tree-sink-stats
-fno-show-column -S  -o ssa-sink-1.s    (timeout = 300)
pid is 14375 -14375
waitres is 14375 exp6 0 0
output is
Checking pattern "sparc-*-sunos*" with i686-pc-linux-gnu
Checking pattern "alpha*-*-*" with i686-pc-linux-gnu
Checking pattern "hppa*-*-hpux*" with i686-pc-linux-gnu
Checking pattern "sparc-*-sunos*" with i686-pc-linux-gnu
Checking pattern "alpha*-*-*" with i686-pc-linux-gnu
Checking pattern "hppa*-*-hpux*" with i686-pc-linux-gnu
ERROR: gcc.dg/tree-ssa/ssa-sink-1.c: error executing dg-final: couldn't open
"ssa-sink-1.c.t46.sink ssa-sink-1.c.t52.sink": no such file or directory
Error/Warning threshold exceeded:  1 0 (max. 1 3)

To solve this problem, the scan-tree-* functions need to check for this case and
remove the old files, so that at least future runs will not suffer from this
problem, eg:
Index: scantree.exp
===================================================================
RCS file: /cvs/gcc/gcc/gcc/testsuite/lib/scantree.exp,v
retrieving revision 1.3
diff -u -r1.3 scantree.exp
--- scantree.exp        31 Mar 2005 00:28:49 -0000      1.3
+++ scantree.exp        6 Jun 2005 07:44:44 -0000
@@ -95,6 +95,12 @@
     set src [file tail [lindex $testcase 0]]
     set output_file "[glob $src.t??.[lindex $args 2]]"

+    if { [llength $output_file] > 1 } {
+    remove-build-file "$src.t??.[lindex $args 2]"
+    error "More than one dump file found: $output_file"
+    return
+    }
+
     set fd [open $output_file r]
     set text [read $fd]
     close $fd


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|FIXED                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910

[Bug testsuite/21910] Files from previous runs can make some testcases fail

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-22 18:22 -------
No this is how it is supposed to happen.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21910