[GIL] Patch for regression test failures in GCC 4.4

[Richard Webb <richard.webb@…>]

The GIL regression tests have a number of failures on GCC 4.4, bue to conflicts between gil::at_c and mpl::at_c.

for example:

​http://tinyurl.com/qag3e3

​http://tinyurl.com/qdxwna

The attached patch avoids these problems by qualifying some calls to at_c with gil::, and additionaly replaces the 'using namespace std' in pixel.cpp with 'using std::swap' in order to avoid an ambiguity with std::prev (as seen in the second link above).

[Hailin Jin]

Fixed in Revisions 54725 and 54726

[peter@…]

Additional fixes.

[peter@…]

The original patch didn't fix the bug on my system; there were still compilation errors involving mpl. I modified two GIL files in a couple places to refer to gil::at_c rather than simply at_c, as the original patch had done elsewhere. With the modified files, GIL compiles on my system, and code using GIL (such as the GIL examples) works. I have not, however, applied any regression tests.

I don't know why the patch here, which presumably worked for the author, did not fix the problem on my system.

[Richard Webb <richard.webb@…>]

Replying to peter@…:

I don't know why the patch here, which presumably worked for the author, did not fix the problem on my system.

The patch i posted was purely to fix failures in the regression tests (i don't use GIL myself, so i didn't do further tests).

There is also #3619 about this same issue, though in that case the existing change already seems to have fixed the reporters problems.

[chhenning]

(In [58777]) Added gil:: namespace to all at_c calls. This seems necessary for current gcc ( 4.4 and more ) compilers. See #3041 and #3619.

[chhenning@…]

I have tried to reproduce the problem with the current trunk ( 01/06/10 ) and release branch using MinGW 5.1.6 ( gcc 4.4 ). The problem could not be reproduced.

[chhenning]

(In [65907]) Added gil:: namespace to all at_c calls. This seems necessary for current gcc ( 4.4 and more ) compilers. See #3041 and #3619.