[Bug translator/21173] New: odd translator behavior with overloaded functions

["dsmith at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=21173

            Bug ID: 21173
           Summary: odd translator behavior with overloaded functions
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: translator
          Assignee: systemtap at sourceware dot org
          Reporter: dsmith at redhat dot com
  Target Milestone: ---

Since bug #18431, the translator supports overloaded functions (2 functions
with the same name but different parameters). However, in certain situations,
this isn't acting correctly.

Here's a test script:

====
function foo:long(a:long, b:long)
{
        return (a + b)
}

function foo:string(a:long, b:string, c:long)
{
        return sprintf("%s %d", b, a + c)
}

probe begin
{
        println(__inode_vfsmount(1))
        println(foo(1, 2, 3))
}
====

====
stap --compatible=2.9 -vp4 ../src/test.stp
Pass 1: parsed user script and 465 library scripts using
135824virt/44664res/7380shr/37236data kb, in 240usr/50sys/293real ms.
WARNING: Eliding unused function 'foo': identifier 'foo' at
../src/test.stp:1:10
 source: function foo:long(a:long, b:long)
                  ^
semantic error: unknown type in dereference: operator '&' at
/usr/local/share/systemtap/tapset/linux/dentry.stp:93:24
        source:                 head = &sb->s_mounts
                                       ^

Pass 2: analyzed script: 1 probe, 8 functions, 3 embeds, 0 globals using
285440virt/195264res/8548shr/186852data kb, in 1830usr/320sys/2167real ms.
Pass 2: analysis failed.  [man error::pass2]
=====

There are at least 3 problems with that translator output:

- It doesn't warn/error on the fact the 'foo' function variants has 2 different
return values.
- Somehow this causes @cast() type errors to appear elsewhere.
- It doesn't realize that there is no overload that takes 3 longs.

It is possible the translator found the error earlier, but its message was
suppressed and it incidentally resulted in shortcuting some other processing
leading the @cast() or whatever to never be attempted.

Note that if the 'foo' function call is fixed to call either function
correctly, the script compiles just fine without any @cast-type errors. Also
note that I found this while debugging a systemtap.stress/tapset_functions.exp
failure.

-- 
You are receiving this mail because:
You are the assignee for the bug.