From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail4-relais-sop.national.inria.fr (mail4-relais-sop.national.inria.fr [192.134.164.105]) by walapai.inria.fr (8.13.6/8.13.6) with ESMTP id pBEH4FUr009893 for ; Wed, 14 Dec 2011 18:04:15 +0100 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AvwAAM3V6E7B/BfSkWdsb2JhbABDqGGCaAEBAQEJCwsHFAMigXIBAQQBOEABBQsLGAkWDwkDAgECATcBDQYNAQcBAReHXwK3N4N5iBAElHSFTYxe X-IronPort-AV: E=Sophos;i="4.71,353,1320620400"; d="scan'208";a="123372415" Received: from msa01.smtpout.orange.fr (HELO msa.smtpout.orange.fr) ([193.252.23.210]) by mail4-smtp-sop.national.inria.fr with ESMTP; 14 Dec 2011 18:04:10 +0100 Received: from [192.168.1.102] ([83.199.117.71]) by mwinf5d49 with ME id 95491i00M1YWU4b03549gG; Wed, 14 Dec 2011 18:04:10 +0100 Message-ID: <4EE8D70A.1030207@frisch.fr> Date: Wed, 14 Dec 2011 18:04:10 +0100 From: Alain Frisch User-Agent: Mozilla/5.0 (X11; Linux i686 on x86_64; rv:8.0) Gecko/20111105 Thunderbird/8.0 MIME-Version: 1.0 To: Adrien CC: Gerd Stolpmann , Jonathan Protzenko , Martin DeMello , caml-list@inria.fr References: <4EDE33A0.6070004@gmail.com> <1323760512.9833.9.camel@samsung> <4EE711FB.5020602@frisch.fr> <4EE83C26.7090108@frisch.fr> <4EE86D90.6080409@gmail.com> <4EE87976.4030604@frisch.fr> <1323876479.7750.36.camel@samsung> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Caml-list] Some comments on recent discussions On 12/14/2011 04:49 PM, Adrien wrote: > Hmmm, right. But if /usr/bin/gcc is already a symlink, ocaml wouldn't be > able to use it at all... I find it quite weird but I don't have a cygwin box > to test. Well, that's precisely the point: the natural way to use gcc 3 under Cygwin is through symlinks in order to have /usr/bin/gcc links to gcc-3.exe. But then ocamlopt and flexlink cannot see it. A solution would have been to have them look directly for gcc-3.exe (and to fix the change of path for /lib/mingw), but we decided to use this opportunity to switch to gcc 4. > But windows actually has symlinks. Kind of. Starting with Vista and the > corresponding NTFS version. But by default you need to be an administrator > to use them, you can only create a limited number of symlink in a given > folder afaiu, some functions work on the symlink and some on the target > (stat()/lstat()). They have a number of limitations and last time I looked > at them, I found them to be mostly unusable because of their limitations. > > They're one quite big issue I've had for packages on windows: if I > cross-compile a library from Linux, and make a tarball which has a number of > symlink in it. What to do when untarring on windows? Try to create symlinks? > Use hardlinks when possible? Copy the file's contents? Something else? Even if Windows supports kinds of symlink internally, this is a rarely used/exposed features. I think it's a bad idea to rely on them for a packaging system (targeted to "native" Windows users). They would look "foreign" to users, and we should expect a lot of bad support from existing tools. Alain