analyzer issue

I seem to be having problems running the analyzer. (This is bug 7406.)

Here's what I did on a Debian testing machine:

checked out llvm and clang trunk into /one/llvm-svn

in /one/llvm-svn-build ran cmake /one/llvm-svn, then ccmake . and just
changed the install path

ran "make", "make clang-test", "make install"

Then changed my path to include the install bin directory and
/one/llvm-svn/tools/clang/tools/scan-build

clang -v reports
clang version 2.0 (trunk 106386)
Target: x86_64-unknown-linux-gnu
Thread model: posix

Then, in hello-2.6 (unpacked GNU hello 2.6) just ran "scan-build
./configure" and "scan-build make" and the analyzer crashes. But it's
apparently not supposed to and I'm seeing problems that were fixed
months ago.

So can anyone suggest what I might be doing wrong?

I’m having the same issue. I filed bug http://llvm.org/bugs/show_bug.cgi?id=7366 last week. It was introduced in June I think

Andy

bruce.r.stephens@gmail.com writes:

[...]

Then, in hello-2.6 (unpacked GNU hello 2.6) just ran "scan-build
./configure" and "scan-build make" and the analyzer crashes. But it's
apparently not supposed to and I'm seeing problems that were fixed
months ago.

So can anyone suggest what I might be doing wrong?

I wondered whether I had old libraries or something confusing things, so
I did the same kind of thing in a pbuilder-created chroot (so this
contains only the base of Debian unstable and a few things that I
installed). I built using ./configure (just to skip installing cmake)
and I get the same crash as before. I installed and used gcc-4.2 to
build again, but that version also crashes.

So I guess there's some difference with linux+elf that's not visible
with Darwin builds? Isn't anyone else using (or trying to use) analyzer
on GNU/Linux?

Andy Wick <Andy.Wick-RGAZE9FVnntBDgjK7y7TUQ@public.gmane.org> writes:

I¹m having the same issue. I filed bug
http://llvm.org/bugs/show_bug.cgi?id=7366 last week. It was introduced in
June I think

Oh, OK. I saw that bug, but for some reason didn't think it was the
same. However, it might well be (I don't see why I'd have thought it's
not).

[...]

i file same PR as http://llvm.org/bugs/show_bug.cgi?id=7363 :slight_smile:

and close it some days ago because after (at least) rev.106215 there are no error with #include <stdio.h>

Dmitry

2010/6/19 Andy Wick <Andy.Wick@teamaol.com>

Дмитрий Дьяченко <dimhen-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
writes:

i file same PR as
http://llvm.org/bugs/show_bug.cgi?id=7363&lt;http://llvm.org/bugs/show_bug.cgi?id=7366&gt;:\)

and close it some days ago because after (at least) rev.106215 there are no
error with #include <stdio.h>

I'm not sure it's the same (I'm using a more recent version than that).
However, I've chopped the failing case to something much smaller, so
I'll attach that to the bug and hope that it helps.

[...]

clang rev.106390 analyze http://llvm.org/bugs/attachment.cgi?id=5020 w/o errs for me…

[root@wellesberg clang_errs]# clang -v
clang version 2.0 (trunk 106390)
Target: i386-pc-linux-gnu
Thread model: posix

Dmitry

2010/6/19 <bruce.r.stephens@gmail.com>

Дмитрий Дьяченко <dimhen-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
writes:

clang rev.106390 analyze http://llvm.org/bugs/attachment.cgi?id=5020 w/o errs for me...

[root@wellesberg clang_errs]# clang -v
clang version 2.0 (trunk 106390)
Target: i386-pc-linux-gnu
Thread model: posix

Works for me, too. However,
<http://llvm.org/bugs/attachment.cgi?id=5059&gt; crashes it.

[...]