Date: Fri, 13 Jul 2001 10:32:31 +0300 From: "Eli Zaretskii" Sender: halo1 AT zahav DOT net DOT il To: pavenis AT lanet DOT lv Message-Id: <2110-Fri13Jul2001103230+0300-eliz@is.elta.co.il> X-Mailer: Emacs 20.6 (via feedmail 8.3.emacs20_6 I) and Blat ver 1.8.9 CC: djgpp-workers AT delorie DOT com In-reply-to: (message from Andris Pavenis on Fri, 13 Jul 2001 09:13:16 +0300 (WET)) Subject: Re: Comments on GCC 3.0 distribution References: Reply-To: djgpp-workers AT delorie DOT com Errors-To: nobody AT delorie DOT com X-Mailing-List: djgpp-workers AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk > Date: Fri, 13 Jul 2001 09:13:16 +0300 (WET) > From: Andris Pavenis > > > > Is it wise to distribute such a major part of the development > > toolchain that is based on a relatively untested library? For > > example, we've just learned that it will be unable to read any files > > on Windows 2000 due to the FAT32 bit in _open. > > > > There might be other problems and incompatibilities, for example in > > the symlink support. > > I think use of DJGPP on Win2K is in alpha stage now. I don't think we can > rate it higher now That's true, but why make the situation worse by introducing additional bugs? > With non modified djdev203 we should have the same NTVDM crash > problem (on nested DPMI tasks) anyway. I didn't meant to say you should use stock djdev203. I think there are a few patches that you should apply to djdev203 to make GCC 3.0 better, but still stable enough to be trusted. I suggest we discuss what patches to apply to djdev203, to build GCC with the patched library. If you agree, I'm willing to post a list of patches I'd recommend, as a starting point for such a discussion. > I used patched libc to build > gcc-3.0 (patch Eli send to workaround NTVDM crashing problem). This is one of the patches we should IMHO apply to djdev203. It needs to be applied anyway, even if you build with the CVS version of the library, since I didn't yet commit it, pending final evaluation by the person who suggested the idea. > Now we have another problem with opening files under Win2K but as I think > it's not new, only it was not diagnosed correctly earlier It is new with the CVS library; it didn't exist in djdev203, because, as far as ANdrew's analysis shows so far, it was caused by adding the FAT32 bit to the open mode passed to the OpenFile function. > it seems that part of problems of latest versions of RHIDE under Win2K was > due to the same problem : RHIDE was not able to read files (including > djgpp.env) under Win2K Wasn't the latest RHIDE built with the CVS library as well, or at least with FAT32 patches? If not, if RHIDE was built with stock djdev203, we might have another problem on our hands.