sourced files can't be always resolved

Reproduce the shown script and inspect the trap handler, especially the debug output for $_Dbg_bash_command, $PWD, and BASH_SOURCE. Investigate how sourced paths are currently resolved from $_Dbg_init_cwd, $_Dbg_cdir, and $(pwd); done means reliably locating ../lib.bash after it is sourced from another working directory, including a regression test for the example.

Follow up to https://github.com/Trepan-Debuggers/bashdb/issues/50
I think that I have a better understanding now of what's going on.

For this script:

cd ./some-dir
. ../lib.bash
cd ~
function_from_lib.bash

The trap handler invoked when function_from_lib.bash is executed, has this data:

• $BASH_SOURCE[1] == ../lib.bash
• $PWD == $HOME

To manage breakpoints, we have to resolve ../lib.bash to its source file.

But atm we're only attempting to guess its directory by looking into directories${_Dbg_init_cwd}, $_Dbg_cdir and $(pwd).
But files can be sourced from any directory, based on the working directory at the time source is called.

To reliably find the source file of ../lib.bash we need to know the working directory when source ../lib.bash is called, i.e. we need to know that it's /path/to/some-dir.

With debug output added to the TRAP handler, this is what's available when cd some-dir and . ../lib.bash is called:

TRAP: commmand='cd 'some-dir'', PWD=/Work/source/bashdb/tmp/sources, BASH_SOURCE=../lib/hook.sh sources/main.bash ../bashdb
TRAP: commmand='. ../lib.bash', PWD=/Work/source/bashdb/tmp/sources/some-dir, BASH_SOURCE=../lib/hook.sh sources/main.bash ../bashdb

Possible approach: In the trap handler, it may work to detect . and source commands in $_Dbg_bash_command and store the absolute path of ../lib.bash based on the current $PWD. This would involve parsing the bash command and there won't be a complete solution handling all edge-cases. But managing the basic source ./file/path ignored-args... should be possible, I think.

I'm not an expert with trap handlers and it's possible that I'm not aware of other, possible solutions.

@rocky Do you think that this may work? Do you perhaps know a better approach to solve this?