This is the mail archive of the cygwin-developers@cygwin.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Issue with cygwin_daemon merge


Conrad Scott wrote:
"Christopher Faylor" <cgf@redhat.com> wrote:

On Fri, Sep 20, 2002 at 09:22:49PM +0100, Conrad Scott wrote:


I've been worried and confused about my proposed merge of the
cygwin_daemon branch, because some of the files in the branch have a
different name from those in HEAD: in particular, this applies to the
System V IPC header files (ipc.h, msg.h, etc.).

In the branch these are in include/sys with the correct names, while

in

HEAD they are in the main cygwin source directory with mangled names
(cygwin_ipc.h, etc.) -- this was done to avoid confusing

configuration

scripts etc. that would otherwise have seen the headers.

I've come to the conclusion that the best bet (i.e., what would make

my

life easiest) is to have the files under the same names in both

branch

and HEAD, but keep them out of the line-of-fire so that nothing finds

by

mistake.  The problem with the location used by HEAD is that they

don't

end up in the installed directory areas and so make testing, even
locally, rather difficult.

My suggestion is to put these headers in include/cygwin with the

correct

names.
Is the interface working in this merge?  If so, then it seems like
putting

the files in sys is the way to go.

Only the shm interface is available, and until all are working and
cygserver can replace the cygipc package, I'm assuming that we can't
expose any of them.


Otherwise putting them in include/cygwin is ok.

Thanks: it sounds like this is the way to go.  I'm not clear if/how I
can produce a patch that moves files: is it possible?

Or would it be okay for me to go ahead and make the change in HEAD
myself?  The change involves moving the relevant files and editing both
them and a couple of the cygserver* files that include them.
Wouldn't it just be easier to use "-x foo.h -x foe.h -x fum.h" type arguments with cvs diff to generate a patch without these headers? Then you could adjust the includes in the source file(s) in a seperate checkin.

Cheers,
Nicholas



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]