Fix Integer Overflow #2357

[chris-smith-zocdoc]

Fixes #2357

Please let me know if this is what you meant in your comment @guolinke

Where should I add tests for this?

[StrikerRUS]

Please use two-space indent, not 4.

[chris-smith-zocdoc]

Please use two-space indent, not 4.

Fixed. We could add an https://editorconfig.org/ file to automatically configure editor settings for people who use it. Example

[StrikerRUS]

How about simple explicit prohibition of large max_depth values in config as per #2357 (comment)?

// check = <=16

LightGBM/include/LightGBM/config.h

Lines 214 to 216 in b310fb4

	// desc = limit the max depth for tree model. This is used to deal with over-fitting when ``#data`` is small. Tree still grows leaf-wise
	// desc = ``<= 0`` means no limit
	int max_depth = -1;

[StrikerRUS]

We could add an https://editorconfig.org/ file to automatically configure editor settings for people who use it. Example

Thanks a lot! Cool feature! I'm adding this to our feature requests (#2302) near the cpp lint test request.

Refer to #2401.

[chris-smith-zocdoc]

How about simple explicit prohibition of large max_depth values in config as per #2357 (comment)?

// check = <=16

LightGBM/include/LightGBM/config.h

Lines 214 to 216 in b310fb4

	// desc = limit the max depth for tree model. This is used to deal with over-fitting when ``#data`` is small. Tree still grows leaf-wise
	// desc = ``<= 0`` means no limit
	int max_depth = -1;

I'm not super familiar with tuning LightGBM yet, but I don't think we want to limit max_depth to such a small value. In one of my datasets I use a value around 40 and wouldn't be surprised if other people are also using a large value.

I would also consider this a significant change and should be discussed outside the scope of this bug fix.

[guolinke]

@StrikerRUS
I agree with @chris-smith-zocdoc , for leave-wise tree, max_depth could be large.

[chris-smith-zocdoc]

@guolinke @StrikerRUS Would you like anything else added to this pr? Should I add tests or update docs?

[guolinke]

ping @StrikerRUS for the test and doc.

[StrikerRUS]

@chris-smith-zocdoc @guolinke I think that new fatal error message is quite clear and no clarifications in docs are needed.

[chris-smith-zocdoc]

Thanks @StrikerRUS I think this is ready to merge then.

[StrikerRUS]

@chris-smith-zocdoc Thanks a lot!

[StrikerRUS]

@guolinke Can we merge this?