PSPP-BUG: [bug #55101] Test 406 "testing #! construct" in segmenter.at fails in pspp 1.2.0 debian build and in master

[Friedrich Beckmann]

URL:
  <https://savannah.gnu.org/bugs/?55101>

                 Summary: Test 406 "testing #! construct" in segmenter.at
fails in pspp 1.2.0 debian build and in master
                 Project: PSPP
            Submitted by: beckmanf
            Submitted on: Mon 26 Nov 2018 10:32:39 AM UTC
                Category: Syntax Parser
                Severity: 5 - Average
                  Status: Confirmed
             Assigned to: None
             Open/Closed: Open
         Discussion Lock: Any
                 Release: None
                  Effort: 0.00

    _______________________________________________________

Details:

The debian regression fails for arm64 and kfreebsd-amd64

https://buildd.debian.org/status/package.php?p=pspp

I could reproduce the problem on debian/testing (amd64) with the current
git master. Test 406 fails which checks the segmenter.

The asseration that fails is here:

http://git.savannah.gnu.org/cgit/pspp.git/tree/tests/language/lexer/segment-test.c#n158

The reason for the problem is that type is not initialized and in the problem
case it has the value SEG_NEWLINE (uninitialized) which then triggers the
assertion because
at that point of time the n value is 11. Input has the value „#!
/usr/bin/pspp\ntitle mq“

The segmenter.c calls the segmenter_parse_shbang

http://git.savannah.gnu.org/cgit/pspp.git/tree/src/language/lexer/segment.c#n86

which iterates through the 11 characters, does not set type, i.e. type remains
unchanged
and returns the value 11 because eof is true. I am not sure if this is a
regression
bug or a real bug because I do not understand what the proper behavior of the

segmenter is.

When you call it with eof=true, it states that it will never return with -1.
How do
you figure out if a segment has been identified or not? What is the expected
behaviour?

Fritz

Am 26.11.2018 um 02:41 schrieb Ben Pfaff <address@hidden>:

There was a build failure on arm64, in one of the tests. I tried to
reproduce it on a Debian arm64 machine, but the same test passed
there.  Very odd.
On Mon, Nov 19, 2018 at 9:00 AM Ben Pfaff <address@hidden> wrote:






    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?55101>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/