`bdb` subshell issues

[delameter]

Hi. Ran into interesting behaviour of your shell debugger. Test script is:

#!/bin/bash
cd "$(dirname "$(readlink -f "$0")")" || exit 127
echo Testing
echo Testing
echo 123

Line 2 is a common way of loading resources relative to script (at least Im using it that way:)

The issue is: when you are setting/deleting breakpoints while running subshells, your changes to breakpoints are local and will be lost after returning to the original script, at least that's what I guess. Example:

$ bdb test.sh    
bdb> trace
bdb> ba true
bdb> +(test.sh:2): cd "$(dirname "$(readlink -f "$0")")"
bdb> +(test.sh:2): dirname "$(readlink -f "$0")"
bdb> bl
     1  true
bdb> bd 1
# after ^ this there should be no breakpoints
bdb> +(test.sh:2): readlink -f "$0"
+(test.sh:3): echo Testing
# however it persists:
bdb> Testing
+(test.sh:4): echo Testing
bdb> Testing
+(test.sh:5): echo 123
bdb> bl 1
     1  true
bdb> 123

Not sure if thats an issue though.

PS Спасибо, в любом случае скрипт полезный, и с год назад мне бы он ооочень пригодился)

[kt97679]

Hi, thank you for using bdb! I think you nailed the root cause of the observed behavior - it is because of the subshells. In the subshell we inherit context of the parent shell so we have all breakpoints defined before starting subshell. Deleting breakpoint in the subshell context will not change parent context so I fully agree this can be confusing. I'm afraid the only thing that can be done is to document this somewhere. In theory we can detect that we are in the subshell and issue warning while adding/deleting breakpoints, but I'm afraid this will add complexity to the script.