Re: svn commit: r19901 - trunk

"Erik Huelsmann" <ehuels@gmail.com> writes:

> On 6/1/06, Philip Martin <philip@codematters.co.uk> wrote:
>
> But you're asking me to reverse engineer Darwin's dynamic linker?

The commit claims that it fixes a problem on your Linux box as well.

>> The discussion on the dev list just stopped
>> without reaching a conclusion :( If you have "before" and "after"
>> versions of the executable it must be possible to determine what
>> changes -- on Linux at least, I can't speak for Darwin.
>
> Before and after what? Ah! Failing! Ok. Backsearching first failure now.
>
> The above sounds a bit frustrated, and frankly, it is: I want to do
> the homework, I just don't know where it is, what it is or how to find
> it... :-(

What I meant was that you claim this commit fixed a problem that
caused svnsync to load the wrong shared libraries. Therefore without
the fix you have a "before" ELF executable that fails, and with the
fix you have an "after" ELF executable that works. Something in the
executable, or the shared libraries, must have changed. I'd guess
it's either the NEEDED or RPATH fields, and you can view those by
running "objdump -x" on the libraries and executable, but it might be
something else. On Linux you can set the environment variable
LD_DEBUG=libs (use LD_DEBUG=help to see other options) and the linker
will produce detailed information about the libraries that get loaded.
Compare the output from the "before" and "after" executables and see
what has changed.

> Thanks for the help. I'll followup when I find the first non-failing rev.

Can't you simply use your change to get working/failing executables?

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org