Win95: "Corrupt intallation detected." - cause known, solution needed

[Tim Peters]

[Bengt Richter]
> But an md5 check settles it even more definitely ;-).

The fellow was on Windows 95 and couldn't install Python, and didn't sound
like the type who might be running Cygwin too.  Therefore he probably has no
straightforward way to compute an md5 checksum.  If he comes back and says
the size-check matched, we can deal with that then.  But *most* times this
comes up the size-check fails, and then he's on the right path after a few
seconds.

> IMHO any file published for downloading should have the md5 digest and
> size on the download page (size more for estimating download time than
> integrity checking).

Heh.  We do, but the size is published only on the SourceForge download site
while the MD5 is only on the python.org site.  Best of all worlds <wink>.

> As you I'm sure you know, the md5 for Python-2.1.1.exe can be found at
>     http://www.python.org/2.1.1/ page:
>
> 39ef54d3e29ea402c8b693b4f3fedd4a Python-2.1.1.exe

Yes, I compute it and email it to Guido.

> (Of course that doesn't prove that a corrupted file wasn't put
> into the archive.  For belt and suspenders, I guess you could do md5's
> for individual critical files within an archive as well, if you don't
> trust the archiving/dearchiving process. Even that will just help
> identify the point of corruption, but at least we'll know where the
> culprit must be ;-)

Our Windows installer .exe can actually be opened as a data file by WinZip
(or similar tool), and it stores a CRC for every file within it; WinZip can
check them all with its "Test" function; indeed, the "installer corrupted"
message comes from the installer doing these CRC checks on itself, so when
it says "corrupted" that's almost certainly the truth (barring strange
permissions problems on NT and 2K).