[Rcpp-devel] Loading a package using Rcpp Modules results in memory corruption

[Dominick Samperi]

On Mon, Jan 10, 2011 at 12:02 PM, Dirk Eddelbuettel <edd at debian.org> wrote:

>
> On 10 January 2011 at 11:36, Dominick Samperi wrote:
> |
> |
> | On Mon, Jan 10, 2011 at 11:09 AM, Dirk Eddelbuettel <edd at debian.org>
> wrote:
> |
> |
> |     On 10 January 2011 at 11:04, Dominick Samperi wrote:
> |     | On Mon, Jan 10, 2011 at 10:32 AM, Dirk Eddelbuettel <
> edd at debian.org>
> |     wrote:
> |     |     We (somewhat breathlessly) await your patches.
> |     |
> |     | I already submitted a clear explanation of how to fix the problem
> earlier
> |     | in this thread, with examples. This involved a lot of work. The fix
> is
> |     | straightforward (if tedious), and without this fix SEXP's will be
> |     | unpredictably blown away, leading to a crash.
> |
> |     Just let us know the URL of your carefully tested patch.
> |
> |
> | It is not my responsibility to fix your memory bugs. I have
> | contributed the solution. How and when you implement it
> | is up to you, and apps that depend on Rcpp may crash
> | unpredictably until you do.
>
> Nonsense. You are just like a crank standing on a soap box in Hyde Park,
> proclaiming the end of the universe.
>

Exactly what part of this is nonsense? Is the memory corruption that was
observed by Doug, me, and you an optical illusion? Is the fix that I
proposed
and that was confirmed by the three of us a dream?

In the process of tracking down this bug I learned quite a bit about
what Romain has added to Rcpp, and admire his creativity. It is too bad
that he had to team up with somebody who is more politician than
software developer, somebody who does not see the absurdity of
insulting somebody who contributes a fix for a serious memory bug
instead of showing some appreciation, and somebody who actually
tries to make excuses for an obvious memory bug instead of
fixing it.

For those Rcpp users who may have doubts, turn on garbage collection
after every allocation by running gctorture() before running the
functions of your package (assuming that your package depends
on Rcpp, of course). It is highly likely that your functions will
fail. They will also run very slowly. If your functions seem to work
without gctorture() that is because you are lucky. There is
no guarantee that a garbage collection will not invalidate one or
more objects that your program depends on.

If you perform this test please feel free to report your results on
this mailing list.


> You have contributed nothing of value here.  I already debunked your
> so-called fix once before. By endlessly repeating the same claims, you are
>

Just to refresh your memory, the reason we parted ways years ago is
that it became apparent to me that you didn't have much to contribute, and
that
you tended to release software that was untested and buggy (sometimes
with my name attached). Now you are free-riding on the efforts of
Romain, and I suspect that he has similar reservations about your
dismissive attitude regarding quality control and reliability (because he
is obviously a very talented software developer).


> not enhancing your position.  Neither does leaving your one package on CRAN
> in its will-not-even-compile state [1].  How much credibility does that
> give
> you, exactly?
>

My package cxxPack does not compile because YOU "released"
RcppClassic, work that is 99% due to me, as "deprecated," for the benefit of
a
user community that does not exist, without checking with me or
doing any kind of integration testing (apparently an alien concept
for you). Releasing this (deprecated!) package makes no sense
given the existence of the current version of Rcpp, and was done
for reasons that have nothing to do with software quality.

Need I remind you that I had to submit bug reports to fix
Date problems in Rcpp, problems that did not exist in the
"deprecated" software from which this Date functionality
stemmed?


>
> Please put up (patches are welcome), or shut up.
>

You are contradicting your remark above that you have "debunked"
my so-called fix. Why then would patches be needed?

Dominick


>
> Dirk
>
> [1] http://cran.r-project.org/web/checks/check_results_cxxPack.html
>
> --
> Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20110110/d9cadb4f/attachment.htm>