From: chrismac AT midland DOT co DOT nz ("Chris McFarlane") Subject: Re: cygwin.dll 16 Jan 1997 22:51:43 -0800 Approved: cygnus DOT gnu-win32 AT cygnus DOT com Distribution: cygnus Message-ID: <20563509703148.cygnus.gnu-win32@midland.co.nz> Reply-To: chrismac AT midland DOT co DOT nz Comments: Authenticated sender is Original-To: gnu-win32 AT cygnus DOT com X-mailer: Pegasus Mail for Windows (v2.23) Original-Sender: owner-gnu-win32 AT cygnus DOT com ================== Date sent: Wed, 31 Jul 1996 10:03:50 -0700 From: Doug Evans Subject: Re: Static compiling To: Robert_Noth Copies to: gnu-win32 Date: Tue, 30 Jul 1996 17:59:09 -0700 From: Robert Noth Is it possible to compile programs with win32 gcc so that the cygwin.dll is not required? Nope, that's not possible. I agree that it would be nice if it were possible. Making it happen, though, may require some compromises on how well the two work together [which may or may not be completely acceptable in any particular situation]. ================== From: Geoffrey Noer Subject: Re: Minimalist GNU-Win32 To: colin (Colin Peters) Date sent: Thu, 17 Oct 1996 01:05:40 -0700 (PDT) Copies to: gnu-win32, paag, tom, tomb Colin Peters taps on the keyboard: > > Some time ago I mentioned that I was working on getting GCC as > distributed by Cygnus decoupled from the cygwin.dll library so I could > write new, non-UNIX-flavoured Win32 programs using GCC. [...] It is a goal to be able to create Win32 programs using gcc that don't require the cygwin.dll. I am in the midst of integrating Scott Christley's headers into the distribution which should help with making more of the Win32 API accessible. The specs file has been changed since beta 16 -- but I don't know whether we made the same changes or not. I'll have to take a look at what you've done when I get the chance... -- Geoffrey Noer noer AT cygnus DOT com ================== > I was about to write something along the lines of > > "A cygwin.dll ought not to be 3Mb big, because I looked at the DJGPP > executables, and nothing approaching 3Mb is required. The smallest > executable is 2k (djtart.exe), the largest is 230k (ld.exe), but > most executables are more modest. gcc.exe is only 99k, for instance. > Okay, there is also the DPMI to consider - but that weighs in at > a miniscule 26k. So, all in all, the executables have quite a small size." > > but then Jim Balter writes > > > The point is that cygwin provides *unix semantics*. While grep > > may not need much of that beyond filenames, something like bash > > certainly does. There is a lot of code in cygwin.dll. > > which seems a fair point. > > On the other hand, if I want to distribute a binary which stands alone > from the cygnus project, then I'd have to include the whopping 3Mb > dll in the distribution - which is a bit off-putting for a > potential downloader. Then, > if he transfers stuff with floppies (which I do, because I'm using > a Win 3.11 486 to connect to the Internet, but I have a Pentium > Win 95 at home, where I do most of my programming), he's then got the > hassle of splitting files up, and recombining them, etc. etc.. > > Fair enough - maybe cygnus has quite ambitious goals which extend > beyond what I am looking for. My own desire is to have UNIX shell > functionality, and all the binaries that go with it, inside a Win 95 > environment. I don't really want to oust Win 95 - because otherwise > I could just install GNU/Linux. Win 95 does have it's nice applications - > eg. Word - and so can't really be written off. When I create an distribution, > I would like the end user to be able to use it as simply as possible. > Which means he doesn't spend time fiddling about with the installation. > > Now, don't get me wrong. I'm not saying that you CAN'T have dll's, > or that install programs are difficult to write. If I'm writing a > fairly nifty visual application, then I'd expect to have some kind of install > utility to set everything up for the user, including any fancy dlls that are > required. > But on the other hand, if I'm writing a shell utility, I'd like the binary > to be complete in itself. > > > Any university offers courses that will cover the tradeoffs between > > dynamic and static libraries; it seems some readers of this group > > could benefit from such an education. > > Oops - I'm duly chastened. But I'd like to point out that commercial > software for Win95 runs fairly smoothly, whereas stuff I get off the > Internet usually requires the Devil's own cunning to get working. > > > Martin Oldfield writes > > > What happens if we change the name of cygwin.dll to include a (major) > > version number ? Then people can keep cygwin1.dll, cygwin2.dll, &c on > > their machine and run both old and new executables without any > > problems. > > This suggestion has the advantage that we could use our old system > to bootstrap the new system, and default to the old one when things get > tricky. The problem I had in the past was that my beta 17 started to > 'act up'. I started to make an installation of 17.1, but things > got mixed up along the line, and I couldn't sort out the mess. I > ended up deleting beta17. I am now in the position of installing 17.1 > from scratch. I doubt I'm the only one that has had these problems. > One of my floppies contains a read error, currently preventing this. > But that's another story. - For help on using this list, send a message to "gnu-win32-request AT cygnus DOT com" with one line of text: "help".