RoothPath command about memoscope.net HOT 4 CLOSED

Hi,

Thanks for the feedback.

I got the same issue on my computer :(
It's a threading issue so I've added a lock on the data reader to avoid this.
Not very statisfied with this fix as lock is slow and "root path" command calls "CountReferers" a lot of time but it will fix the bug until I find a better solution.

from memoscope.net.

Note for me: during GUI refresh, some modules calls "CountReferers" while the RootPath command is running in another thread and calling the method too.

from memoscope.net.

This is a great project.
I was playing with tables with a sqllite browser.

CREATE INDEX IdxInstanceAddress ON Instances (Address)
CREATE INDEX IdxReferencesRefBy ON InstanceReferences (RefByAddress)

select * from (
WITH RECURSIVE
allreferences(address, level) AS (
values(435582892, 0)
union
SELECT refbyaddress, allreferences.level+1 FROM instancereferences, allreferences
WHERE instancereferences.instanceaddress=allreferences.address and allreferences.level < 5
)
SELECT* FROM allreferences ) r, instances ins , types t where r.address = ins.address and ins.typeid=t.id

But sqllite doesnt allow more complex queries and this doesnt go anywhere with this.
-should run without level
-exclude already included addresses
-differentiate if this is a leaf node
etc

I am gonna try a different approach for finding root path(s) with changing actual sources.
However not sure, how to show them in UI.

This is in my mind for now
-Create a new table CLRROOT (probably simply address column)
-Populate this table in the init stage
-Create RootPath table (not sure but; analyseaddress, level, leaf, address, refbyaddress etc)

Instead of beginning with clrroots, begin with just the address of given object, go for referers until a clrroot found.
This way it should be more faster and also all roothpaths can be found.
(our clrroot objects are like 2.3M)

from memoscope.net.

btw, i had a few other sqllite errors before, those looked like some concurrent usage atm.
(while rootcommand is running and as it takes too much time, I tried to do run some other)

I am taking a look at source code, I think each command should have its own con/trx.
If i am not mistaken they are using ClrDumpCache con/trx with could cause concurrency problems with sqllite

from memoscope.net.