Le 07/06/2016 08:54, Jean-Marc Lasgouttes a écrit :
Le 05/06/2016 01:01, Guillaume Munch a écrit :
4-11. Replace Boost features with std equivalents when possible. The
result is more consistency across plaforms and fewer dependencies on
Boost. More details down below. For your reading convenience, patches
1-11 are also attached together in batch0.diff.
This is OK for me (if the extra srd:: are removed).
I was a bit fast here.
4-6. Straightforward and without risk.
Indeed. When this is over, we shall update our boost source tree to
remove unused stuff.
7-8. These are essentially examples. While there is no hurry to apply
these patches, they are a good example of how the unique_ptr introduced
in the first commit let us simplify the class definitions. I believe in
"do not fix the unbroken", but I also believe that the whole source
would benefit from being less convoluted.
Sure, it is good to use modern C++11 constructs. I may complain at some
time that it breaks my oldish gcc 4.6 setup, but we will see what to do
if/when this happens.
Actually, I will try to apply and compile your whole patchset with my
ancient compiler.
9. There should be no bad surprise with <cstdint> itself, however since
the purpose of everything in strfwd.h is not clear to me, I would rather
have a second opinion.
I do not know much about that either. I suspect we will have to abandon
strfwd.h at some time.
10. boost::bind and boost::function. For this one it would be preferable
to have a confirmation that it is safe. In particular, some files used
the boost version explicitly even when lyx::function was already defined
from the standard library. Is there a particular reason?
I think there is no real reason.
11. boost:regex. We have learnt the pitfalls of having two regex
libraries at the same time. However, according to a comment in the
source, gcc < 4.9.0 supports std::regex badly. Do we still care about
gcc < 4.9.0 ? (no hurry, and the patch is trivial)
As was written by others already, it is too early. Ubuntu 14.04 still
has 4.8.2.
JMarc