#1 Re: mORMot 2 » TSynDaemon/TSynDaemonSettings memory leak » 2024-11-11 07:19:30

Hi

I actually did put in a TSampleDaemon.Free() and got the exception as you said.

So then thought is was either of the two above (Sorry AB)

Like I said, there are brighter people out there.

Thank you very much.

#2 mORMot 2 » TSynDaemon/TSynDaemonSettings memory leak » 2024-11-10 20:13:07

rudiloos
Replies: 4

Hi

Busy writing a little server and wanting to put on a Linux machine.
I already have 2 daemons running that was compiled around 2.2.7700 on the main server but never checked for leaks.
Not sure if they have leaks when they are shutdown as they are compiled as releases.
However they also show leaks when compiling them with report memory leaks.

For simplicity I then used the standard Project05Daemon sample to confirm, and this is what these results are based on.

For testing I am running the server with the /verbose and --verbose parameters.
I assume that is what it is there for.

Keep on getting memory leaks for the above mentioned items.
Upgraded to 2.3.8854 (Today) and the results are the same.
Mormot2Tests runs clean.

Leaks for Win32, Win64, Linux 64 with similar outputs. (one 6 blocks the other 5 blocks etc)

Tried to step through the code but the leak only appears after the server actually shuts down
and I am unable to pinpoint where. The log file shows no errors.
Clearly this is above my capability.
A lot more brighter people out there might be able to shed some light.

Windows 10, FPC 3.2.3, Laz 3.4
Ubuntu 24.04 LTS VM with all updates done (Did not test on main server)

CLI outputs for Win32 and Linux64
https://gist.github.com/rudiloos/6b684f … 57781b4b51

Thanks

#4 Re: mORMot 1 » Batch columns and values mismatch » 2022-11-18 19:27:00

It does shed some light and I appreciate the effort, thank you.

The TRecordVersion, as you say, is not a simple field but as I understand it, only for for retrieval. Posting and updating should be transparent.
The main concern for me is when the record count is changed. That should have no bearing on the fields.

As I said, there is no rush and I will wait until ab is back from his well deserved vacation.

#5 Re: mORMot 1 » Batch columns and values mismatch » 2022-11-14 14:51:32

Thanks ab,

Thanks for the reply.
I have just created a dummy field for now.
Still have a lot of other parts to keep me busy.

No worries. Enjoy your holiday.

#6 mORMot 1 » Batch columns and values mismatch » 2022-11-14 00:41:14

rudiloos
Replies: 9

Hi

The following code

https://gist.github.com/rudiloos/dcbc66 … a0edb7989f

produces the following error

Error SQLITE_ERROR (1) [insert into Test (RowID,CreatedOn,CreatedBy,EditedOn,EditedBy,Notes,RecStatus,File_Name,File_Type,File_Ext,File_Size,File_Date,File_MD5,ProjID,DoneOn,DoneBy,COCNo,COCType,SentOn,SentTo,DateBPM) values (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)] using 3.39.4 - 22 values for 21 columns, extended_errcode=1

Looking at the fields, it seems to not list the TRecordVersion field but wants to insert the value.

This only happens when the record count added is 199. No error if the record count is 198 or 200.
Changing the model by increasing or decreasing the field count by one also does not generate an error.
Adding the records one by one and not in batch mode produces no errors.

Tried to look through the code but way above my intellect and just gave me a headache.

Am I doing something wrong or is it a bug?

Synopse mORMot framework 2.0.4191, compiled with Free Pascal 3.2.2 32 bit, against SQLite 3.39.4, on Windows 10 64bit (10.0.19045)

Board footer

Powered by FluxBB