Clarification on the support of process substitution in OpenSSL

[hrafnthor]

I have be unable to use process substitution with OpenSSL when doing so via bash while, on the same system, having no issues running the same command with zsh.

The command in question is the following (with a valid credential replacing mock value):

echo "something to sign" | openssl dgst -binary -sha256 -sign <(echo "credential")

When run via bash (verified via ps -p $$) the command will fail with error:

Could not read private key from /dev/fd/63

Again I stress, the same command above works flawlessly when run using zsh.

Now the question becomes can this behaviour be attributed to OpenSSL it self, or would this be due to some quirk in the implementations of the shells?

I ask since there are numerous examples online showing OpenSSL being used with process substitution, indicating that for some this clearly does work when used with bash, but for others it does not work at all (no matter what the shell is).

---

System information

Ubuntu 22.04 - 5.19.0-46-generic
OpenSSL 3.0.2 15 Mar 2022 (Library: OpenSSL 3.0.2 15 Mar 2022)
zsh 5.8.1 (x86_64-ubuntu-linux-gnu)
GNU bash, version 5.1.16(1)-release (x86_64-pc-linux-gnu)

[hrafnthor]

It feels very silly now but it turns out that the differences in behaviour were due to backslashes in the credential payload, and bash requiring interpretation of backslash escapes to be turned on via the -e flag (i.e using ... <(echo -e "credentials").

[mspncp]

It's not silly at all. Instead, it should feel satisfying that you were able to solve your riddle all by your own. :-)