Bug 20672 - @defined(@cast()) regression
Summary: @defined(@cast()) regression
Status: RESOLVED FIXED
Alias: None
Product: systemtap
Classification: Unclassified
Component: translator (show other bugs)
Version: unspecified
: P2 normal
Target Milestone: ---
Assignee: Unassigned
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-10-07 13:30 UTC by Martin Cermak
Modified: 2017-01-26 13:18 UTC (History)
3 users (show)

See Also:
Host:
Target:
Build:
Last reconfirmed:
Project(s) to access:
ssh public key:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Martin Cermak 2016-10-07 13:30:32 UTC
On rhel6, buildok/dentry-embedded.stp regressed.  The new error is:

=======
spawn stap -p4 --compatible=2.9 /root/systemtap/testsuite/buildok/dentry-embedded.stp
semantic error: while processing function real_mount

semantic error: type definition 'mount' not found in 'kernel': operator '@cast' at /usr/local/share/systemtap/tapset/type_defined.stpm:6:16
        source:     @defined(& @cast(0, @type) -> @member)
                               ^
        in expansion of macro: operator '@type_member_defined' at /usr/local/share/systemtap/tapset/linux/dentry.stp:160:6
        source:         if (@type_member_defined("mount", mnt_parent)) {
                            ^

Pass 2: analysis failed.  [man error::pass2]
wait results: 12735 exp6 0 1
FAIL: buildok/dentry-embedded.stp
=======

The regression seems to be caused by following commit:

=======
commit 056cb27baac1ce3ab4d675dbbe4881afde801ca3
Author: Frank Ch. Eigler <fche@redhat.com>
Date:   Wed Jun 22 11:43:33 2016 -0400

    PR18079: support nested autocast / @defined
    
    We now perform const-folding & dead-code-elision during the type
    resolution loop, whenever an autocast expression gets evaluated.  This
    way, @defined(foo()->mm) type expressions can work as nature intended.
    
    This requires @defined() not to be short-circuit evaluated to 0 during
    a random const_folding process, so a flag is introduced to control its
    preservation or collapsing.  For the last (assert_resolvability) pass
    in the type resolution loop, this flag is set to true, so that
    genuinely unresolvable @defined($expressions) do get mapped to 0 in
    time for a last elision.
=======
Comment 1 Martin Cermak 2017-01-06 12:43:45 UTC
Another problem this aforementioned commit 056cb27baac1 is causing is, that it breaks the compileability of sys.stp (part of syscall.exp) on rhel6:

=======
 6.8 Server x86_64 # stap -I $HOME/src/testsuite/systemtap.syscall/tapset  -p4 $HOME/src/testsuite/systemtap.syscall/sys.stp
semantic error: while processing probe kernel.function("sys32_mmap@arch/x86/ia32/sys_ia32.c:155") from: syscall.mmap32 from: syscall.*

semantic error: type definition 'mmap_arg_struct32' not found in 'kernel': operator '@cast' at /usr/local/share/systemtap/tapset/linux/x86_64/syscalls.stp:169:15
        source:         if (@defined(@cast($arg, "mmap_arg_struct32")->addr)) {
                                     ^

Pass 2: analysis failed.  [man error::pass2]
 6.8 Server x86_64 #
=======
Comment 2 Frank Ch. Eigler 2017-01-20 20:55:01 UTC
Just for the record, this is not exactly a regression, but the exposure of a latent bug:

probe kernel.function("sys_sendto").return { println(@defined(@entry($foobar))) }

fails under quite old systemtap too.
Comment 3 Frank Ch. Eigler 2017-01-25 15:53:05 UTC
monster commit 16aa72d9adf4 reworks some rewriting passes
Comment 4 Tetsuo Handa 2017-01-26 04:26:32 UTC
Hello.

This commit is causing a regression (presumably infinite loop)
with python probes.

---------- /tmp/test.py start ----------
def func1(a, b):
  return a + b;

print(func1(1, 2));
---------- /tmp/test.py end ----------

# time PYTHONPATH=/tmp /usr/bin/stap -e 'probe python2.module("test").function("*") { printf("%s %s\n", pn(), $$parms); }' -c 'python -m HelperSDT /tmp/test.py'



Bisect log:

# bad: [e81970274b46cf3e20586751bf58195258f31518] Add timestamp to nfsd-trace.stp example.
# good: [9f950728d1a921e32d53cc8dabcbc0684d0256ee] Update lookup_bad_addr() to avoid a kernel warning.
git bisect start 'HEAD' '9f950728d1a921e32d53cc8dabcbc0684d0256ee'
# good: [4fb0437d6e19bc5b7c917af5f542938d68968828] testsuite: more adaptation to $var -> @entry($var) in .return probes
git bisect good 4fb0437d6e19bc5b7c917af5f542938d68968828
# bad: [16aa72d9adf420e973cfb3370f4cb9dacd3922fd] PR20627: rework several code rewriting passes
git bisect bad 16aa72d9adf420e973cfb3370f4cb9dacd3922fd
# good: [15528f73d31f4fdd5257016957ac2e4cfaac8cd0] testsuite: tweak bz1027459
git bisect good 15528f73d31f4fdd5257016957ac2e4cfaac8cd0
# first bad commit: [16aa72d9adf420e973cfb3370f4cb9dacd3922fd] PR20627: rework several code rewriting passes
Comment 5 Frank Ch. Eigler 2017-01-26 13:18:06 UTC
This appears to be fixed in commit f90352b.
It was caused by interesting infinite recursion of the python
variable-expander in-place mapping 

$var3
to
python2_get_var_obj(Py2Object_Repr($arg3, "arg3"))

and then the nested $arg3 ad infinitum.