msvcr100.dll missing ... error started after Windows 10 update to 10586.17

Christian Gollwitzer auriocus at gmx.de
Sat Dec 5 04:07:20 EST 2015


Am 05.12.15 um 00:26 schrieb Glenn Linderman:
> My wife's 64-bit Win8 home machine has 32-bit Python 3.3 installed.
>
> Then it upgraded to Win 8.1. Then I upgraded it to Win 10. Then I
> upgraded it to Threshold 2. It gets regular automatic updates also, like
> the one last night to build 10586.17.
>
> That's the history.
>
> When she tried a python script today, it failed, with an error saying
> that MSVCR100.dll was missing.
>
> After a few false starts, like being surprised that the error happened
> when it worked yesterday, and that there was an MSVCR100.dll in
> %windir%\system32, doing a search for all MSVCR100.dll on her machine
> discovered quite a few in various application directories, but then also
> one in \windows.old\WINDOWS\SysWOW64, the light-bulb blinked on, I
> copied that one to the \python33 directory, and everything works.

These MSVCR*DLL are a bit different from other DLLs. They constitute the 
runtime for programs compiled using Visual Studio 2010.  Instead of 
mucking around with these files manually, you should install the 
"Redustributable Package" found here

https://www.microsoft.com/en-us/download/details.aspx?id=14632
for 32bit or here
https://www.microsoft.com/en-us/download/details.aspx?id=5555
for 64 bit. Both can be installed simultaneously to support 32bit and 
64bit programs.

There is also a "Visual Studio 2010 SP1" version - I'm not sure which 
one is correct. It depends on the version of the compiler that Python 
was built with.

> Why M$ chose to delete MSVCR100.dll from %windir%\SysWOW64 in the recent
> update is a mystery, however.

Maybe the upgrade process didn't recognize it as part of the 
rediistributable, maybe it was not installed correctly. The MS 
recommended way for an installer is to package that "redistributable" 
and to launch it during the installation of the main program.

> So this is just a data point and warning and solution, not really an
> expectation that anyone will be able to explain M$.
>
> Glenn




More information about the Python-list mailing list