Re: [xml] runtest mystery bug: name2.xml error case regression test

[Daniel Richard G. <oss teragram com>]

On Tue, 11 Sep 2012, Daniel Veillard wrote:

>  Really I tried again today to make sure, and even with an out of tree 
> build and your patch applied, it just work as expected for me on Fedora 
> 17 x86_64,
>
> thinkpad:~/tmp/build -> ./runtest
> ## Error cases regression tests
> Total 11 tests, no errors
> thinkpad:~/tmp/build ->
>
>  I can try 20 times it's always the same,

I got a VM of FC17 up and running, and the test passes for me every time 
there too. Try as I might, I can't make it fail.

At this point, I'm not sure *what* I have. The test fails consistently on 
my work desktop (Ubuntu 11.10), yet if I build on an 11.10 live desktop 
with the same compiler, it works. If I copy over the binary from my work 
system to the live desktop, it fails. But if I copy that same binary to 
the FC17 VM, it works. (Note that other systems at my office, with older 
versions of Ubuntu, give the same error; it's not just my machine.)

If there were a good way of getting an execution trace, I could do a diff 
between what's happening on my system versus on FC17. Doing "step" in a 
loop in GDB doesn't work, however, returning clearly bogus results (e.g. a 
call to __xmlRaiseError() that doesn't occur in a normal debugger run).

This will have to be left unresolved for now, until/unless I can get a 
better handle on it :-(


--Daniel


--
Daniel Richard G. || danielg teragram com || Software Developer
Teragram Linguistic Technologies (a division of SAS)
http://www.teragram.com/