| Message ID | 201007030608.o6368Ltk023530@ignucius.se.axis.com |
|---|---|
| State | New |
| Headers | show |
On 07/02/2010 11:08 PM, Hans-Peter Nilsson wrote: > I don't see the patch for that commit on the gcc-patches@ nor > fortran@ lists. > > Anyway, the patch had a formerly hidden (due to the invalid > dg-syntax you fixed) syntax error; an unquoted ')', causing: We discussed the dg-syntax problems on IRC. There was another one broken after fixing the syntax that I found, but I mussed this one. Thanks for fixing. Jerry
Index: gcc/testsuite/gfortran.dg/char_bounds_check_fail_1.f90 =================================================================== --- gcc/testsuite/gfortran.dg/char_bounds_check_fail_1.f90 (revision 161765) +++ gcc/testsuite/gfortran.dg/char_bounds_check_fail_1.f90 (working copy) @@ -10,4 +10,4 @@ zz(i:j) = 'abcdef' print * , zz end -! { dg-output "Substring out of bounds: upper bound exceeds string length.*at line 9)" } +! { dg-output "At line 10.*Substring out of bounds: upper bound \\(11\\) of 'zz' exceeds string length" }
I don't see the patch for that commit on the gcc-patches@ nor fortran@ lists. Anyway, the patch had a formerly hidden (due to the invalid dg-syntax you fixed) syntax error; an unquoted ')', causing: Running /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/dg.exp ... ERROR: tcl error sourcing /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/dg.exp. ERROR: couldn't compile regular expression pattern: parentheses () not balanced while executing "regexp $texttmp ${output}" invoked from within "if ![file exists $output_file] { warning "$name compilation failed to produce executable" } else { set status -1 set result [${tool}_l..." (procedure "saved-dg-test" line 224) invoked from within "saved-dg-test /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/char_bounds_check_fail_1.f90 { -O0 } { -pedantic-errors}" ("eval" body line 1) invoked from within "eval saved-dg-test $args " (procedure "dg-test" line 9) invoked from within "dg-test $test $flags ${default-extra-flags}" (procedure "gfortran-dg-runtest" line 27) invoked from within "gfortran-dg-runtest [lsort \ [glob -nocomplain $srcdir/$subdir/*.\[fF\]{,90,95,03,08} ] ] $DEFAULT_FFLAGS" (file "/tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/dg.exp" line 32) invoked from within "source /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/dg.exp" ("uplevel" body line 1) invoked from within "uplevel #0 source /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/dg.exp" invoked from within "catch "uplevel #0 source $test_file_name"" Not only that, but the old text was wrong; the diagnostic apparently has changed. AFAICT, the test should pass. The diagnostic emitted when testing for cris-elf is: At line 10 of file /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gfortran.dg/char_bounds_check_fail_1.f90 Fortran runtime error: Substring out of bounds: upper bound (11) of 'zz' exceeds string length (10) So, I believe it was correct and obvious to commit the following patch. You may want to follow-up and remove the exact bound specifications which weren't present in the old text, but please be careful about the regexp syntax. Of course running the test-suite after the change and inspect the log helps. ;) BTW, those regexp evaluations really should be in some kind of "catch"-clause in whatever lib/*.exp, so they don't mess up the rest of the test-run! (Which is how I spotted it; a FAIL was unexpectedly gone.) Or perhaps reported to bug-dejagnu if it can't be fixed like that locally. Commited after observing that the above was fixed. gcc/testsuite: * gfortran.dg/char_bounds_check_fail_1.f90: Correct dg-output string. brgds, H-P