Re: Compilation speed on Windows

[Robert Ögren <gtk roboros com>]

On Tue, 8 Mar 2005, Tor Lillqvist wrote:

> I wrote:
>  > This is on a quite fresh XP SP2 machine, with recommended security
>  > fixes, and very little third-party software yet installed
>
> Hmm, forgot to mention Symantec AntiVirus... which brings up the
> quesiton, could it be your antivirus (or maybe anti-adware) software
> that causes the slowdown for you? (Even if it for some reason shows up
> for Explorer?) One might think that anti-adware software could install
> some hooks to look especially closely at any "installers" being
> run. Although, one wonders how long it can take for adware writers to
> get the hint and start calling their installers something else than
> install.exe...

The machines run antivirus software (all from different vendors) and are
up to date XP SP 2.

Thanks to WinDbg and the publicly available Windows debug symbols I was
able to get a better understanding of the issue. There is a function,
SHELL32!CDesktopBrowser::_SetupAppRan , that is called after an
installation program has terminated. Just attach WinDbg to the
explorer.exe process, point it to Microsoft's public symbol server, set a
breakpoint at the function, and run any program called install.exe.

Using Regmon it is possible to see that one of the things it does is loop
over all(?) registered COM classes and look for those that implement a few
component categories (for explorer toolbars I think), and create a cache
with the classes it found. On a machine with some programs installed there
are a bunch of COM classes registered, so this may explain some of the CPU
time that is taken. Perhaps this is why you don't see much CPU usage on a
fresh machine, or maybe there are other factors involved. The wonders of
closed source software...

Anyway, I'll stop investigating this now, it seems to be intended
behaviour. Sorry for the offtopicness.

Robert