Mailing-List: contact cygwin-apps-help AT sourceware DOT cygnus DOT com; run by ezmlm Sender: cygwin-apps-owner AT sourceware DOT cygnus DOT com List-Subscribe: List-Archive: List-Post: List-Help: , Delivered-To: mailing list cygwin-apps AT sources DOT redhat DOT com Message-ID: <3B8A5FAB.1030202@ece.gatech.edu> Date: Mon, 27 Aug 2001 10:56:43 -0400 From: Charles Wilson User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.2) Gecko/20010713 X-Accept-Language: en-us MIME-Version: 1.0 To: cygwin-apps AT cygwin DOT com Subject: Re: Updated setup.ini with descriptions, categories, and dependencies References: <20010827004327 DOT A14852 AT redhat DOT com> <3B89E74D DOT 3060705 AT ece DOT gatech DOT edu> <20010827025736 DOT A15753 AT redhat DOT com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit > > Thanks. > > cgf > Christopher Faylor wrote: > On Mon, Aug 27, 2001 at 02:23:09AM -0400, Charles Wilson wrote: >>cvs: >> cgf: ash cygwin perl tcsh zlib >> >> comments: well, perl isn't actually *necessary* for cvs itself, it's >>only used by a few additional scripts. Also, tcsh is not required at >>all. OTOH, the cygwin build of cvs requires gdbm. And crypt. And >>'pr.exe' which is part of textutils. Also, since the cvs source >>includes its own version of zlib, it doesn't really depend on the >>external zlib package. >> > > I wonder why the RPM version is so skewed from what you are relaying here. well, there is a rational explanation for *most* of it: 1) gdbm: you can build cvs with gdbm as an option. I chose to use that option; I suppose Red Hat did not choose to do so. 2) zlib: there's probably a config option for "use platform zlib" or some such. I have not specified that option (if it exists) so my cvs build uses the internal zlib code. 3) crypt: the crypt() functions are built in to the linux kernel, so you don't really need an external package 4) tcsh, textutils: ??? > > However, if some part of cvs needs perl to operate then this is a dependency. As far as splitting cvs into two packages, I don't have a problem with that. (I've been wondering if we shouldn't split some of the library packages into, for instance, readline- and readline-devel- (or even "libreadline4" (containing only cygreadline4.dll and cyghistory4.dll), "libreadline5" (containing only cyg<*>5.dll and lib<*>5.a), "readline" (containing the executables and docs), and "readline-devel" (containing only the import libs and header files). This new "readline" package and "readline-devel" package are un-major-versioned because you can only have ONE set of those installed at a time, corresponding to the most recent libreadline package. But that's an awful lot of work; one thing at a time...) > > >>jpeg: >> cgf: --- >> csw: cygwin >> (category should be Graphics/Libraries, not miscellaneous) >> > > This is a difference from RPM. >> sdesc: "A library of functions for manipulating JPEG image format files" >> > > As is this. Right. The RPM (a) doesn't specify an sdesc at all, and (b) is just plain wrong about the category. Jpeg is not miscellaneous -- it's a graphics library. Now, since the jpeg source code contains both the library and a few applications, perhaps Red Hat split it up into "libjpeg" (Graphics/Libraries) and "jpeg" (miscellaneous, contains the executables) IMO, even *that* is marginal; I'd put BOTH into Graphics/Libraries. > > My script didn't properly translate the jpeg -> libjpeg but I have rectified that. > > I've made all of the changes you've mentioned. I still wonder why RPM got some > of them "wrong". I hate to let you in on this secret, but RPMs -- Red Hat as well as 3rd party -- routinely get dependencies wrong. Usually by artificially requiring a higher version that is ACTUALLY required, forcing unneeded upgrades, but also by just plain leaving stuff out or including additional unnecessary "dependencies". And of course, categorization is just a matter of opinion, anyway. --Chuck