[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: g++ 3.0 patch for apt (partial)

first, apologies to the list for not sending that diff gzipped.... it
was kinda big..

> Some of the stringy stuff should probably stop using the char* ranges and
> use just strings, conversion overhead be damned. (Like the Hex2Num
> function!)

your call, but i'd prefer not to touch that for now.

> I'm confused about the changes to pkgcachgen.cc, the pkgIndexFile vector
> iterator should be used throughout the entire pkgIndexFile list interface,
> but it is not.

there probably was a reason, but i don't remember right now.

> We talked about the policy.cc thing, that looks like it still needs a
> solution.. 

yeah, vectors of structs and all that.

> ios::nocreate.. hmmm.. Presumably ios::in implies that, but that has to be
> confirmed, both for GCC 3 and GCC 2.95

should be easy to do.

> Is adding 'using [..]' acceptable practice for a header file?  Should all

my understanding was that "using foo:bar" is ok but obviously "using
namespace std" would be bad in a .h

> of those be moved within the class scopes? My C++ books are all so dated
> that they only have descriptions of pre-ISO namespaces.. 

i'll investigate this some more.

> Wot is this:
> -       sed -e "1s/.*:/$(subst /,\\/,$@):/" $(basename $(@F)).d > $(DEP)/$(@F).d
> +       sed -e "1s/.*:/$(subst /,\\/,$@):/" $(DEP)/$(basename $(@F)).d > $(DEP)/$(@F).d
> How can that be correct? Did GCC3 do something horrid to the -MD flag?

um, yeah, it did weird things with the paths. this is just a hack for

> Hmm:
> +ostream c0out(0);
> +ostream c1out(0);
> +ostream c2out(0);
> I wonder if that makes it segfault.

seems to work.

> I need to cut this up into smaller pieces, and have a gcc 3 place to test
> on.. Did taggart ever get gluck setup?

not yet. g++-3 is on auric tho, but hppa is probably tracking gcc cvs a
bit more closely than the official debs.

i have some more cleanups to get apt-ftparchive to work, will send those
along at some point too. perhaps we can do a g++3 branch or something?
i'd like to get these into cvs asap so we can build official hppa apt debs.

incidentally, unless the g++ maintainer changes the naming convention my
libc/libstdc++ versioning stuff won't work correctly with the current g++3
libstdc++ file naming convention.

Debian Developer <tausq@debian.org>

Reply to: