[Bash bug] Building BaCon on MSYS2

[ghost]

The problem was found by @sskras: msys2/MSYS2-packages#3960 (comment)

[sskras]

Printing CR and LF characters using echo -e works as expected in all tested environments:

• Cygwin 3.4.6-1 (2023-02-14)
• Midipix 2022.11.18
• MSYS2 3.4.6 (2023-04-01)
• Ubuntu 20.04.06 LTS

But when echoing CR-LF during command substitution (recording output of the command between the backticks), in some cases MSYS2 behavior starts to differ:

What is echoed	Simple echo, generic case is OK	During command substitution		MSYS2 state
		on Cygwin 3.4.6-1, Ubuntu 20.04, midipix 2022.11	On MSYS2 3.4.6
CR-only, no LF:	$ echo -e -n "\015" | od -tx1 0000000 0d 0000001	$ echo -n `echo -e -n "\015"` | od -tx1 0000000 0d 0000001	$ echo -n `echo -e -n "\015"` | od -tx1 0000000 0d 0000001	OK
CR + implicit LF:	$ echo -e "\015" | od -tx1 0000000 0d 0a 0000002	$ echo -n `echo -e "\015"` | od -tx1 0000000 0d 0000001	$ echo -n `echo -e "\015"` | od -tx1 0000000	Broken
CR + explicit LF:	$ echo -e -n "\015\012" | od -tx1 0000000 0d 0a 0000002	$ echo -n `echo -e -n "\015\012"` | od -tx1 0000000 0d 0000001	$ echo -n `echo -e -n "\015\012"` | od -tx1 0000000	Broken
Leading CR-LF:	$ echo -e -n "\015\012." | od -tx1 0000000 0d 0a 2e 0000003	$ echo -n `echo -e -n "\015\012."` | od -tx1 0000000 0d 20 2e 0000003	$ echo -n `echo -e -n "\015\012."` | od -tx1 0000000 0d 20 2e 0000003	OK
Middle CR-LF:	$ echo -e -n ".\015\012." | od -tx1 0000000 2e 0d 0a 2e 0000004	$ echo -n `echo -e -n ".\015\012."` | od -tx1 0000000 2e 0d 20 2e 0000004	$ echo -n `echo -e -n ".\015\012."` | od -tx1 0000000 2e 0d 20 2e 0000004	OK
Trailing CR-LF:	$ echo -e -n ".\015\012" | od -tx1 0000000 2e 0d 0a 0000003	$ echo -n `echo -e -n ".\015\012"` | od -tx1 0000000 2e 0d 0000002	$ echo -n `echo -e -n ".\015\012"` | od -tx1 0000000 2e 0000001	Broken

I would summarize the issue following:

Bash on MSYS2 eats trailing CR-LF bytes during the command substitution (instead of eating just the LF byte).

[sskras]

Ping @lazka: do you (or someone of other core devs) agree with my summary (the last line) ?

[sskras]

The behavior is the same on MSYS2 3.4.8.x86_64 2023-08-18 23:11.

[ghost]

Ping @lazka: do you (or someone of other core devs) agree with my summary (the last line) ?

It seems no one cares.

[sskras]

Ping @lazka @elieux. We've got a bug in the build of Bash at handling CRs. Cygwin build is fine.
Maybe we should keep this issue open (even if you are not responding) ? :)

[sskras]

Also ping @MehdiChinoune.

[lazka]

Ok, even if no one has time right now. Let's keep this open.

[sskras]

I could do some R'n'D about the Cygwin-vs-MSYS2 difference. But I have no idea about how to do it / where to start.

[sskras]

Still present:

$ uname -a
MINGW64_NT-10.0-19044 DESKTOP-O7JE7JE 3.4.10.x86_64 2023-12-22 10:06 UTC x86_64 Msys