Guido Serassio wrote:
Hi Amos,
Sorry, I missed your direct question ... :-(
At 06:31 23/04/2008, Amos Jeffries wrote:
Guido:
I have seen the squid_mswin.h include currently in the 'OS-specific
macros' section.
Wrapping the code that include can be moved down at least to above the
types section (done). It also looks like it duplicates many of the
general
definitions made in squid_types.h. Can you confirm that it works if moved
after the squid_types.h include and tweaked?
If I remember right, into squid_mswin.h there are only the needed type
definition.
And there are no duplicate definitions:
inttypes.h and sys/bitypes.h are not available on Windows.
I'm not sure about what happens moving it.
I think worst case might be some re-definition errors. But we don't want
those either. I'll trust you on that and do the move...
The structure of squid_types.h (unchanged by me) is
- include every types* file known to man and autoconf
- test for and define every individual type squid uses that may
sometimes not be defined by all the above.
Most of the types files are unique to a bunch of OS and not available
elsewhere. Windows is not unique here. The POSIX type definitions done
for windows fall into that definition. The 'sometimes' being on windows.
With squid_mswin.h following squid_types.h we should be able to
seamlessly push any typedefs and defines up into the basic squid_types.h
config without side effects.
I think that before changing anything here, we should use the same type
detection used in Squid2 configure, where the missing types are defined
directly into autoconf.h.
Interesting prospect for the long term. Right now I'm only dealing with
.h and .cc. Makefiles and configure are further down the checklist.
One thing I am worried about there is with length of configure.in
already being quite cumbersome. Alex mentioned some sort of auto-tools
include method that may spread hope but I have not researched any of
that yet.
Amos
--
Please use Squid 2.6.STABLE19 or 3.0.STABLE4
