[sr #110985] "Substitute and define special characters" test fails on NetBSD 9

[Zack Weinberg]

URL:
  <https://savannah.gnu.org/support/?110985>

                 Summary: "Substitute and define special characters" test
fails on NetBSD 9
                   Group: Autoconf
               Submitter: zackw
               Submitted: Wed 20 Dec 2023 01:59:11 PM UTC
                Priority: 5 - Unprioritized
                Severity: 2 - Minor
                  Status: Need Info
                 Privacy: Public
             Assigned to: None
        Originator Email: 
             Open/Closed: Open
         Discussion Lock: Any
        Operating System: None


    _______________________________________________________

Follow-up Comments:


-------------------------------------------------------
Date: Wed 20 Dec 2023 01:59:11 PM UTC By: Zack Weinberg <zackw>
On x86_64-unknown-netbsd9.3, running Autoconf's test suite consistently
produces this failure:


Failed tests:
 NUM: FILE-NAME:LINE     TEST-GROUP-NAME
 270: torture.at:913     Substitute and define special characters


This test checks whether it is possible to pass a string through AC_SUBST that
includes a byte with every possible value except 0x00.

The log for the testsuite says


configure: creating ./config.status
config.status: creating Foo
config.status: creating Zardoz
config.status: creating config.h
./torture.at:987: cat Foo
./torture.at:1000: cmp allowed-chars Zardoz
--- /dev/null   2023-12-19 14:19:58.701932966 -0500
+++
/net/u/1/z/zackw/autoconf-2.72d.16-9800e-dirty/tests/testsuite.dir/at-groups
/270/stdout     2023-12-19 14:19:58.910181000 -0500
@@ -0,0 +1 @@
+allowed-chars Zardoz differ: char 256, line 2
./torture.at:1000: exit code was 1, expected 0


Comparing hex dumps of the files "allowed-chars" and "Zardoz" shows that the
problem is that an extra byte (with value 0x81) has been inserted before each
byte with a value in the range 0x81 through 0x8b.  The extra bytes are *not*
present in the value of @zardoz@ recorded in config.status, but they *are*
present in the temporary file `subs1.awk` used by config.status (to see this
file run `./config.status -d` and then look in the directory named confXXXXXX,
where XXXXXX is some random sequence of alphanumerics).  Thus, this appears to
be a bug in the handling of here documents in this system's /bin/sh.  However,
the obvious cut-down version of the code in config.status does _not_ show the
bug:


$ printf 'xxd
<<\\EOF\n\x80\x81\x82\x83\x84\x85\x86\x87\n\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f\nEOF\n'
> test.sh
$ /bin/sh test.sh
00000000: 8081 8283 8485 8687 0a88 898a 8b8c 8d8e  ................
00000010: 8f0a                                     ..


We need to pin down exactly what the bug is before we can do anything about
it.

If you encounter a failure in this test on any platform _besides_
*-unknown-netbsd9*, please post a comment here.  Please attach a tarball of
the directory tests/testsuite.dir/270 (note: the number may be different if
you are testing a version of Autoconf other than 2.72).






    _______________________________________________________
File Attachments:


-------------------------------------------------------
Date: Wed 20 Dec 2023 01:59:11 PM UTC  Name: failed-test-270.tar.gz  Size:
133KiB   By: zackw

<http://savannah.gnu.org/support/download.php?file_id=55471>

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/support/?110985>

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