Ever since upgrading to Resharper 5.0, the performance of Visual Studio 2010 has basically been non-usable for myself and my team. I own a copy, but my team members just evaluated it. It starts out working fine, but then after ~60 minutes of normal use without memory usage skyrocketing (devenv.exe going from 250M on startup to over 800MB when it finally becomes non-responsive).
Per other threads, we've tried upgrading to v5.1 and I just tried v5.1.1. v5.1 seemed a little better, but v5.1.1 was not (it actually seemed worse). However, if we use VS2010 without Resharper enabled, there's no problem. I tried to capture a couple minutes of activity using dotTrace, but once we started profiling devenv.exe basically became entirely unresponsive (CPU went from 3-4% to almost 100%). I uploaded the file to your ftp server - snapshot-jal.zip
FYI - here are the steps I can follow to quickly cause Resharper to kill VS2010:
1) Open medium-sized ASP.NET MVC solution (~25 projects, 500,000 lines of code, .NET 4.0)
2) Pull a tab to a second screen
3) F5 to debug the app
4) Edit an aspx page while the app is running
5) Reload the page that was just edited
This is the fastest way to hit the 800MB point where VS2010 crashes. However, even if I never debug an app and never use multi-monitors, the same thing happens, but takes a little longer (2-4 hours before a restart is needed). Without Resharper, VS2010 runs fine ... one of our developers uninstalled it after his eval and hasn't shut down VS2010 in over a month (vs. twice daily VS2010 restarts with Resharper).
Is there anything we can do? We'd love to use Resharper for our entire team, but at the moment it's just not working.
I also uploaded a dotTrace file while I performed the exact same actions without Resharper running (snapshot-jal-no-resharper.zip). With Resharper, it took 2 minutes, without, 15 seconds. Also, with Resharper, dotTrace took over 20 minutes to load up VS2010, without Resharper active dotTrace took 2 minutes to load Resharper. I realize that for the test much of the time was due to the profiling overhead (all I did was move a tab and replace a "." with another "." to activate intellisense), but the relative difference matches our perceived experience as well. Any help would be greatly appreciated, we REALLY want to be able to use this product.