X-Recipient: archive-cygwin AT delorie DOT com X-SWARE-Spam-Status: No, hits=-1.0 required=5.0 tests=AWL,BAYES_50,J_CHICKENPOX_63,J_CHICKENPOX_83,J_CHICKENPOX_93,RCVD_IN_DNSWL_LOW,SPF_HELO_PASS,SPF_PASS X-Spam-Check-By: sourceware.org To: cygwin AT cygwin DOT com From: "Akakima" Subject: Re: side effects after installing gcc-3.4.4.999 Date: Mon, 2 Mar 2009 21:47:08 -0500 Lines: 153 Message-ID: References: <49AB8757 DOT 90807 AT gmail DOT com> <49AC2ECA DOT 7040709 AT gmail DOT com> Reply-To: "Akakima" Mime-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit X-IsSubscribed: yes Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: cygwin-owner AT cygwin DOT com Mail-Followup-To: cygwin AT cygwin DOT com Delivered-To: mailing list cygwin AT cygwin DOT com "Dave Korn" wrote: > Windows will do because that's outside my control. (And I don't > think it > would make sense to avoid using Linux features in Cygwin just so > that stuff > will work from DOS!) I agree 99.99 % with you. :-). the 0.01 is about would you please not totally forget that cygwin is running on a Win32 platform. (no offense here, no joke, no disrecpect, ...) Technicaly it's not DOS, but the WinXP command prompt window. This windows behave like a DOS console but this is not a DOS application. This a win32 application using a rather feature limited subset of the winapi. (sorry if you already know this). > Wouldn't a good solution be for your windows path to have an extra > directory > in it ahead of Cygwin's /bin directory, where you could put a > windows shortcut > to gcc-3.exe and call it "gcc"? That should work; by having this > directory at > the front of your PATH in DOS (but not in Cygwin) you could easily > add > overrides for the names of any Cygwin applications/symlinks that you > wanted. Yes, it is a good idea. I already have a set of batch files that fix the PATH an other environments variables depending with what compiler i am working with. setcygwin.cmd for Cywin setmingw.cmd for MingW etc... This works quite well. Each new instance of the cmd.exe can have its own set of variables. > Yes, or it could just be a version difference. I use win2k on my > main system. I use WinXP Pro, SP2. > Something very strange is happening! GCC doesn't read any input > from stdin > in any of those cases, so it's surprising that it behaves > differently. Ah! > It might be easier than you think to change them all in one go: > find . -name '*.bat' | xargs sed -i -e 's/gcc.exe/gcc-3.exe/g' Thanks for the tip. > > OK, hang on a moment, I know it might sound silly of me to ask but > now I > have to check: you aren't doing something unusual like running > cmd.exe in a > rxvt window are you? Or do you by any chance have CYGWIN=tty set in > your No. No. > environment variables? I wonder if it's really "waiting for an > enter", or if > it's actually waiting for a full command line, and just the prompt > hasn't been > flushed to the screen yet. What happens if you type "dir " > or some > other command instead of just pressing enter on its own? Now you got it. After doing some other test that's what i found. Effectively if you type a command there, that command is executed. So this is cmd.exe waiting for input (but the prompt has not been displayed or has been erased) > > In your earlier examples, where I said "Cygwin doesn't read stdin > anyway", > that's still the case. But there is one difference that might be > the reason > here: when you run "gcc file.c", gcc launches a number of > subprocesses - > preprocessor, compiler, assembler, linker. With the other two > commands, it > runs and exits without launching any other executables; I wonder if > somehow > that means that in one case all the output gets flushed and you see > the prompt > and know it's finished, in the other case the prompt gets lost in a > buffer > somewhere and you don't see it. Maybe launching subprocesses causes > all the > buffers to be flushed through in the first case. I dont know and i dont know how to test that. I made some other tests. I created the shortcut from the explorer. May be cygwin was creating the shortcut in a way that caused that behaviour, or may be it was because cygwin does not use \r (just \n). Same results. I also tried to create a shortcut to mingw gcc.exe (which emits \r\n). Same results. And i tried with a small program who does only write to stderr, and compiled it with VisualC++ v6.0 : Same results. Conclusion: this is related to the way cmd.exe process and run the target of the shortcut. Shortcuts are normally ran from the explorer. I guess bash is a little more intelligent. > I would advise you consider my suggestion of having a > renaming-and-overriding directory full of windows shortcuts at the > head of > your PATH. There are quite a lot of things in /bin that exist > primarily as > symlinks to the real application, gcc is just one among many. That would be perfect, because with all the good settings, the change would be completly transparent. If cmd.exe was not the culprit. As illustrated. > Hey, I just tried that out to see if it works, and it does, but it > also > allows me to see the "waits for enter" problem occurring! It's > definitely > just the DOS prompt getting lost somehow; any command that you type > before > pressing enter does get executed when you do. Ah, look, it's not > that the > prompt is getting lost: it's just that GCC's output comes out after > the prompt > for some reason. See the example below: > Is that how it's happening for you? Yes!. Exactly. I tried many variants to launch the shortcut. Each time, the results are the same. Thanks for taking the time to verify this. If i find a solution i will report it here. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/