yes, seeing the erranous line where the wrong formatstring comes from is nice but sometimes (for myself: often) i need to know what actually was tried to be logged. this information is lost in both cases.

this is especially sad if you have long-running processes that encounter an error after a long time and that error is logged but the logging itself fails due to formatting issues. i suggest to log whatever is avaiable in the suggested way to at least see this valuable information. the additional information where it comes from for further fixing is apreaciated as well. btw: the stackstrace of the logging internals could be left out in my oppinion, it just pollutes the output with no means to the programer.