Apply max_log_file constraint when timestamp_in_log_file=false

[andrembcosta]

Currently, when the flag timestamp_in_log_file=false and max_log_file is set, instead of quitting logging or overwriting to preserve the size limit, glog just appends to the log file and eventually breaks the size limit.

Am I missing another flag to get the intended behavior of having a single log file of limited size? Or is this an actual limitation?

This appears to be similar to #809

[HNOONa-0]

Hello @drigz,

I’ve investigated this bug and understand why it occurs. I would like to work on fixing this issue
But let's assume that I restrict the file to 1 MiB and I have exceeded that limit, I have the following options:

• Force timestamp_in_log_file to true. Note this can lead to duplication errors when file size constraint is too small (1 MiB)
• Add an auto-incrementing index to LogFileObject and start writing to a new file named base_filename_.<index>
• Truncate the current file once it exceeds the limit and start writing from the beginning

I assume that approach 3 works because if the original intention of the user was to keep multiple files he would use a timestamp.

[drigz]

@HNOONa-0 I only have context on the Bazel integration so maybe you want @sergiud to look at that comment.