This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: winmm.dll error 487 is easily reproducible

On 4 March 2011 09:52, Corinna Vinschen wrote:
> On Mar Â4 05:09, Andy Koppe wrote:
>> On 2 March 2011 13:08, EXCOFFIER Denis wrote:
>> > P.S.: By the way, what is the normal way to install snapshots? It is
>> > easy to replace cygwin1.dll (keeping around the working one), but in order
>> > to test the cygwin1 package in full (eg to check within stdio.h whether
>> > vsnprintf is included under __STRICT_ANSI__ or not)?
>> You should usually install the cygwin-inst-*.tar.bz2 tarball
>> containing all the relevant headers and tools along with the DLL.
>> There's a FAQ entry describing how to install it at
>> The snapshots page could do with a link to that. However, due to
>> Cygwin 1.7's ability to overwrite in-use files, including the Cygwin
>> DLL itself, it's no longer necessary to exclude the DLL when
>> unpacking, i.e. I find the following works for me:
>> - Shut down all Cygwin processes, including background services.
>> - Open a Cygwin shell (with administrator rights if Cygwin was
>> installed as administrator).
>> - cd /
>> - tar xvf /path/to/cygwin-inst-*.tar.bz2
> Just for the records:
> At this point you just have to keep in mind that the old DLL has now
> been moved to the trash bin. ÂEvery new process started from here, by
> fork or by exec, would load the new DLL. ÂThat would very likely collide
> with the memory layout of the old DLL so you're risking a crash. ÂYou're
> lucky that the processes up to here don't fork during un-taring...
>> - Close shell.
> that's the only sane thing to do here.

To avoid this sort of manual faffing about and increase the chances of
people testing snapshots, how about delivering them as test releases
through setup.exe, with announcement to cygwin-announce? Not every one
of them, especially early in a development cycle, but I think it would
make sense when trying to knock a release into shape.


Problem reports:
Unsubscribe info:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]