Date: Wed, 26 Mar 1997 13:01:41 +0300 (IDT) From: Eli Zaretskii To: Teun Burgers cc: djgpp AT delorie DOT com, Marc Singer Subject: Re: ANNOUNCE: RCS version 5.7, patch 13 In-Reply-To: <9703251203.AA16673@joost.ecn.nl> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Tue, 25 Mar 1997, Teun Burgers wrote: > On the networked drive (in this case a DEC-ultrix file system mounted by > DEC pathworks) rename fails if the target file already exists. So it's a libc bug after all. Can we please have such bugs reported here, rather than worked around in the application code? There should be no reason for such work-arounds in the case of DJGPP, since the sources for everything are freely available. The whole point of free software is that correcting library bugs and sharing the information about them is much easier than with proprietary libraries, where you depend on the good will of the vendor. Instead of introducing work-arounds for what seems to be mysterious failures, let us fix them, so others won't need to stumble upon them in the future. Since I don't have access to the particular configuration where the bug happens, I need to ask you to investigate this further, in order to understand the reason for this failure. Please try this test program and tell if it works in that case: #include main (int argc, char **argv){ printf("renaming %s to %s\n",argv[1],argv[2]); printf("return code: %d\n",_rename(argv[1],argv[2])); } (note that I used `_rename' rather than `rename'). In future, when you see such bugs, please don't silently work around them. If you can get the sources, debug the problem and suggest a patch, fine; if you can't, at least report the problem and describe how it can be reproduced so others will know about the bug and fix it when they have time.