Re: [PATCH] cygcheck -s should not imply -d
On 13/01/2011 12:33, Corinna Vinschen wrote: On Jan 11 14:26, Jon TURNEY wrote: On 11/01/2011 08:10, Corinna Vinschen wrote: I wasn't quite sure either, but while running cygcheck with Jon's patch it started to make more sense. We can also change the docs to ask for `cygcheck -svrd' output, but I guess we should just wait and see. FWIW (I don't have all packages installed), mutt is the only package I have installed for which cygcheck -c falsely reports a problem. $ cygcheck -c | grep -v OK Cygwin Package Information PackageVersion Status mutt 1.5.20-1 Incomplete Do you happen to know why? You can read my ill-informed speculation about this matter at [1] :-) [1] http://sourceware.org/ml/cygwin-apps/2010-11/msg00065.html Would a patch to http://cygwin.com/setup.html be welcome recommending that: (a) if a package installs files which a user is expected to customize, don't trample over those customizations when the package is upgraded/reinstalled Isn't that what /etc/defaults and /etc/postinstall is for, basically? I'm not sure I understand what you're proposing. At which point should setup warn and how is it supposed to know that a file is a user-customizable one? In theory, that's all in the responsibility of the package. Sorry, that URL isn't very helpfully named. I'm not proposing to change setup.exe, I'm just suggesting adding some text to the 'Cygwin Package Contributor's Guide' web page, recommending those things. (I only became aware of the existence of /etc/defaults by looking at what other packages do, I can't see it mentioned on that page) (b) a package should verify as correctly installed with cygcheck -c? I don't understand this, sorry. Would you mind to rephrase and maybe give an example what you mean?
Re: [PATCH] cygcheck -s should not imply -d
On Jan 13 13:04, Jon TURNEY wrote: On 13/01/2011 12:33, Corinna Vinschen wrote: On Jan 11 14:26, Jon TURNEY wrote: On 11/01/2011 08:10, Corinna Vinschen wrote: I wasn't quite sure either, but while running cygcheck with Jon's patch it started to make more sense. We can also change the docs to ask for `cygcheck -svrd' output, but I guess we should just wait and see. FWIW (I don't have all packages installed), mutt is the only package I have installed for which cygcheck -c falsely reports a problem. $ cygcheck -c | grep -v OK Cygwin Package Information PackageVersion Status mutt 1.5.20-1 Incomplete Do you happen to know why? You can read my ill-informed speculation about this matter at [1] :-) [1] http://sourceware.org/ml/cygwin-apps/2010-11/msg00065.html Uh, ok. Thanks for the pointer. Would a patch to http://cygwin.com/setup.html be welcome recommending that: (a) if a package installs files which a user is expected to customize, don't trample over those customizations when the package is upgraded/reinstalled Isn't that what /etc/defaults and /etc/postinstall is for, basically? I'm not sure I understand what you're proposing. At which point should setup warn and how is it supposed to know that a file is a user-customizable one? In theory, that's all in the responsibility of the package. Sorry, that URL isn't very helpfully named. I'm not proposing to change setup.exe, I'm just suggesting adding some text to the 'Cygwin Package Contributor's Guide' web page, recommending those things. (I only became aware of the existence of /etc/defaults by looking at what other packages do, I can't see it mentioned on that page) Ouch. Sorry about that. Yes, sure, it would surely be welcome to see progress in the docs, too. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader cygwin AT cygwin DOT com Red Hat
Re: [PATCH] cygcheck -s should not imply -d
On Thu, Jan 13, 2011 at 01:33:36PM +0100, Corinna Vinschen wrote: On Jan 11 14:26, Jon TURNEY wrote: On 11/01/2011 08:10, Corinna Vinschen wrote: I wasn't quite sure either, but while running cygcheck with Jon's patch it started to make more sense. We can also change the docs to ask for `cygcheck -svrd' output, but I guess we should just wait and see. FWIW (I don't have all packages installed), mutt is the only package I have installed for which cygcheck -c falsely reports a problem. $ cygcheck -c | grep -v OK Cygwin Package Information PackageVersion Status mutt 1.5.20-1 Incomplete Do you happen to know why? I know why. It just isn't high on my list of things to fix. cgf
Re: [PATCH] cygcheck -s should not imply -d
On 11/01/2011 08:10, Corinna Vinschen wrote: On Jan 10 12:52, Christopher Faylor wrote: On Mon, Jan 10, 2011 at 01:51:02PM +0100, Corinna Vinschen wrote: On Jan 5 19:50, Jon TURNEY wrote: Currently, for cygcheck -s implies -d. This seems rather unhelpful. I'm afraid I've lost the thread which inspired this, but in it the reporter provided cygcheck -svr output as requested, but this did not help diagnose what ultimately turned out to be the problem, that a DLL was actually an older version (presumably due to replace-in-use problems) Attached a patch to modify cygcheck so -s no longer implies -d (although -d can still be used). 2011-01-05 Jon TURNEY * cygcheck.cc (main): don't imply -d from -s option to cygcheck Looks good to me. Applied. Sorry that I didn't reply to this. I wasn't 100% convinced that this was a good idea since some of the packages show up as having problems when they are ok. I was wondering if that would end up generating more (understandably) confused mailing list traffic but I guess, in the end, it probably is better to check the validity of the packages for the prescribed error reporting technique. I wasn't quite sure either, but while running cygcheck with Jon's patch it started to make more sense. We can also change the docs to ask for `cygcheck -svrd' output, but I guess we should just wait and see. FWIW (I don't have all packages installed), mutt is the only package I have installed for which cygcheck -c falsely reports a problem. $ cygcheck -c | grep -v OK Cygwin Package Information PackageVersion Status mutt 1.5.20-1 Incomplete Would a patch to http://cygwin.com/setup.html be welcome recommending that: (a) if a package installs files which a user is expected to customize, don't trample over those customizations when the package is upgraded/reinstalled (b) a package should verify as correctly installed with cygcheck -c?
Re: [PATCH] cygcheck -s should not imply -d
On Jan 5 19:50, Jon TURNEY wrote: Currently, for cygcheck -s implies -d. This seems rather unhelpful. I'm afraid I've lost the thread which inspired this, but in it the reporter provided cygcheck -svr output as requested, but this did not help diagnose what ultimately turned out to be the problem, that a DLL was actually an older version (presumably due to replace-in-use problems) Attached a patch to modify cygcheck so -s no longer implies -d (although -d can still be used). 2011-01-05 Jon TURNEY * cygcheck.cc (main): don't imply -d from -s option to cygcheck Looks good to me. Applied. Thanks, Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader cygwin AT cygwin DOT com Red Hat
Re: [PATCH] cygcheck -s should not imply -d
On Mon, Jan 10, 2011 at 01:51:02PM +0100, Corinna Vinschen wrote: On Jan 5 19:50, Jon TURNEY wrote: Currently, for cygcheck -s implies -d. This seems rather unhelpful. I'm afraid I've lost the thread which inspired this, but in it the reporter provided cygcheck -svr output as requested, but this did not help diagnose what ultimately turned out to be the problem, that a DLL was actually an older version (presumably due to replace-in-use problems) Attached a patch to modify cygcheck so -s no longer implies -d (although -d can still be used). 2011-01-05 Jon TURNEY * cygcheck.cc (main): don't imply -d from -s option to cygcheck Looks good to me. Applied. Sorry that I didn't reply to this. I wasn't 100% convinced that this was a good idea since some of the packages show up as having problems when they are ok. I was wondering if that would end up generating more (understandably) confused mailing list traffic but I guess, in the end, it probably is better to check the validity of the packages for the prescribed error reporting technique. cgf
