Paradox 10 hanging since upgrade to Windows 10

Hi,

Has anyone managed to get Paradox 10 working satisfactorily on Windows 10? Care to share your setup please?

We upgraded to Windows 10 last week and since then Paradix hangs 4-5 times a day. We manage to get it going again by having everyone log off and deleting all the lock files but it's getting a but frustrating.

We have our data in a shared folder and everyone is pointing to the same .NET file. I've tried every suggestion I've seen here and elsewhere with no luck so far.

Any tips, pointers, suggestions gratefully received. 

Thanks 

Delordson

in general, p9/10/11 run fine under win10

is there a pattern to the problem? specific process? specific tables? specific users?

have you done the ordinary table repair tests, to see if there are any underlying problems?

properly shared net file, cfg file, shared tables, etc., should work fine

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

Hi Steven, 

Thanks for coming back on this and good to know it shouold work.

We are trying to identify a pattern but have not spotted any. I've done table verification on the big tables and spotted one that needed to be re-built but I haven't checked each and every table. Is there a systematic way to do the 'ordinary table repair tests' for all tables? 

The net, cfg and shared tables are all in a network folder and have not changed.

The only thing I can think of, the upgrade to Windows 10 was done on a single PC, Paradaox was installed and checked to be working fine, then that PC was cloned to all other PC's. Not sure if reinstalling on all individual PC's might help?

Thanks

Delordson

paradox should be installed on each machine.. not sure how complete the registry entries are by cloning.. or how many of the registry entries you're avoiding by using a good cfg file, either..

upgrade or not, patterns can show that a particular table has issues.. patterns can show that printing might be an issue.. patterns can show that specific users are creating issues

we are also finding generic issues with SMB2 vs SMB1, and the differences in how they handle oplocks and caching.. this could be a biggie, but the information comes in slowly now

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

We'll uninstall and re-install on all PC's and report back.

I'll do some more detective work to see if we can stop a pattern. Is there a log somewhere I shoul look through?

Thanks

Delordson

sometimes, the "detective work" requires a serious understanding of the app and its' structure

we had an issue on one app recently, where paradox would hang when the user did a specific common search.. there was a table linked to the main table on a sec index.. the sec index in the linked table was hosed.. most searches on that field were ok, but that specific common search caused an "out of range" error

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

Hi Steven,

Can I ask please, how did you detect the broken sec index? Was it literally by checking what everyone was doing when the app froze? 

I was planning to run Verify on all tables but I've read somewhere that this can pass even when the table is broken under some cercumstances?

yes, Verify can miss things.. run Verify first.. look at the Errors table in the priv, it gives a little more detail than the regular screen.. then Rebuild

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

Try this,  Create a new Script,  paste the contents of the two functions into the script and run.


isTableCorrupt example

method run ( var eventInfo Event)

var

var
tbName String
alldata, ErrList String
Tlist Tcursor
endVar

gettables()
ErrList = ""

If TList.Open("_tempTabList") then
While TList.eot() = False
tbname = Tlist.TableName
Try
message(tbname)
if isTableCorrupt (tbname) then
ErrList = Errlist +tbname +" is corrupt\n"
endif

OnFail
ErrList = ErrList + "tbName" + " - ERROR\n"
EndTry
tList.NextRecord()
EndWhile

Tlist.close()
if errlist = "" then
Msginfo("FINISHED","No Errors found")
Else
Msginfo("FINISHED",ErrList)
endif
endif

endMethod

method getTables()
var
  dbName, fileSpec, tbName String
  tv1 TableView
endVar

; Init variables.
  dbName = ":WORK:"
  fileSpec = "*.db" ; Lists <filename>.db
  tbName = "_tempTabList"

enumDatabaseTables(tbName, dbName, fileSpec)

;tv1.open(tbName) ; Open the created table.

endMethod

I highly recommend running that like this

istablecorrupt(tablename, tablename+" errors")

so you can look at the results.. it's just running Verify with the built-in Errors table.. the results can be meaningful

first, ignore all the "low ascii" stuff, that's NOT meaningful.. it matters, over time and over the batch, if you're getting errors in the indexes or in the tables

let us know what you see

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

Thanks Rum, Steven,

isTableCorrupt looks a very useful function. Embarrased I hadn't come across it before. I only started doing Paradox a couple of years ago when I inherited this database from the previous head of my lab. We are looking to migrate the database application to ASPNETCore but that needs a bit more time.

I'll try these suggestions when I'm back at work on Monday and let you know.

OK so I ran the script and it was helpful (see below for how I ended up running it in the end). There were two table with corrupt secondary indexes. I fixed these. It hasn't completely solved the problem unfortunately. I think we may have a network performance issue since we went to Windows 10. Anyone have experience of slow networks with folder that intermittently drop off? I think this may be contributing to our issues. I've read here or somewhere else that that alias needs to be mapped to the UNC rather than the drive letter. Any thoughts on this please?

method run ( var eventInfo Event)

var

                databaseName string

                fileSpec                                                string

                tableName                         string

                alldata                                   string

   errList                                                string

                tablesCursor      tcursor

endVar

                databaseName = ":WORK:"

                fileSpec = "*.db"

                tableName = "_tempTabList"

                enumDatabaseTables(tableName, databaseName, fileSpec)

                errList = ""

                tableName = ""

                if tablesCursor.Open("_tempTabList") then

                                while tablesCursor.eot() = False

                                                tableName = tablesCursor.TableName

                                                Try

                                message(tableName)

                                                                if isTableCorrupt (tableName, "ErrorsIn" + tableName) then

                                                                                errList = errList + tableName + " is corrupt\n"

                                                                endif

                   OnFail

                                                                errList = errList + "tableName" + " - ERROR\n"

                EndTry

                                                tablesCursor.NextRecord()

                                endWhile

                                tablesCursor.close()

                                if errList = "" then

                                                Msginfo("FINISHED","No Errors found")

                                else

                                                Msginfo("FINISHED",errList)

                                endif

                endif

endMethod

drive letters are fine, as long as they're all the same

if it's dropping out, check for network cards that have "sleep" modes, machines that go into "sleep" mode, etc

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN

Have you checked the auto-disconnect setting?  

To turn-off the auto-diconnect feature, you could also edit the Windows Registry. Navigate to the following key in the registry:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\lanmanserver\parameters

Change the hexadecimal value of autodisconnect to ffffffff. Click OK and Exit.

Alternativey, you could also download and apply this Fix It from KB297684.

Followup question

Is there a programmatic way to use ObjectPal to do a table repair?

Chimney Sweep.. but for most folks, I say that you don't want an automated routine cuz it makes it more likely that you won't actually look at the results after the repair.. REAL damage will show easily, via bad number and date values and/or missing records

Steven Green
Senior Software Engineer
Paradox Support Specialist DOS/WIN