tesseract: Disable Leptonica messages

[stweil]

Disable debugging and informational messages from Leptonica
for release builds.

Signed-off-by: Stefan Weil [email protected]

[amitdo]

#522

stweil commented

... without those nagging messages, I'd never have had a look on that part of Leptonica.

DanBloomberg commented

Alternatively, for a bit more flexibility, you can define the environmental variable
#define LEPT_MSG_SEVERITY L_SEVERITY_WARNING
and use
setMsgSeverity(L_SEVERITY_EXTERNAL);

[amitdo]

I want the above option in the release build.
@stweil?

[stweil]

Using the environment variable would imply that you always get at least one message from Leptonica because the code which enables / disables warnings also writes messages. :-( That's why I decided against that feature for the moment.
Leptonica needs a modification which allows suppressing messages from setMsgSeverity, no matter whether the environment variable exists or not and no matter which message severity level is set.

[amitdo]

cc: @DanBloomberg

[DanBloomberg]

First, some more of the basics, for disabling at least all info and warning messages.

You can define the external env variable to L_SEVERITY_ERROR. That will disable warnings, which I believe makes sense for tesseract.

If you set this severity level at compile time, you guarantee that it can't be over-ridden at run-time.

If you want to go further and suppress all messages, including those from the ERROR_* macros (which are different from the informational L_* macros), define NO_CONSOLE_IO on the compile line (or use L_SEVERITY_NONE).

As for the suggestion that setMsgSeverity() tells you what it does, I can easily add a 2nd arg (debug) and only emit those messages when debug = TRUE. Stefan, is that what you want?

[DanBloomberg]

Better, I'll just comment out the comments, leaving the interface unchanged.

[stweil]

Yes, I'd prefer keeping the interface unchanged and disabling the messages (at least for arguments in the valid range).

[DanBloomberg]

this change is pushed to leptonica master on github

[DanBloomberg]

and modified to keep the warning(error) message but not the info(OK) messages

[stweil]

An unset environment variable should not result in a warning, as this is the normal case. With the latest Leptonica code this requires unnecessarily tricky code. I'd use L_INFO instead of L_WARNING.

[DanBloomberg]

I was thinking that if you try to set it to the external value, and that value is not set, it should issue a warning because the attempt failed. I am already commenting the L_INFOs out there, and I can comment the warning out as well.

[DanBloomberg]

All leptonica messages in setMsgSeverity() are now disabled, as requested.

[egorpugin]

Can we change this to #ifdef NDEBUG ? NDEBUG is more common than DEBUG.

[stweil]

Either DEBUG or NDEBUG is set by configure (depending on --enable-debug). It looks like cmake currently does not set DEBUG. Maybe that's the reason why api/baseapi.cpp also tests for _DEBUG (set my the MS compiler).

NDEBUG is a standard macro which changes the behavior of assert. That's also the reason why it is used rather often.

If we want to separately control standard features and Tesseract features, we need both macros. Otherwise we could stick to using NDEBUG.

Personally I prefer keeping both macros, but I think that removing DEBUG would also work.