Fixup for "Add quiet mode for bootstrap"

[Matthias Köppe]

#30533 broke ./bootstrap in unconfigured source trees.

---

References:

​https://unix.stackexchange.com/questions/65532/why-does-set-e-not-work-inside-subshells-with-parenthesis-followed-by-an-or

[Matthias Köppe]

I can just change it back to using #!/usr/bin/env bash...

[Matthias Köppe]

New commits:

​f537959	Trac #30533: add new -q (quiet) flag for the bootstrap process.
​6547b33	Trac #30533: hide automake (stderr) output from ./bootstrap -q.
​171b947	Trac #30533: improve ./bootstrap usage() a bit.
​7dcb5a8	Trac #30533: whitespace cleanup in ./bootstrap.
​ab83893	Trac #30533: rename bootstrap-download() to bootstrap_download().
​2b4272d	Trac #30533: use /bin/sh to run the bootstrap script instead of bash.
​b85f4a1	Trac #30533: hide output from "./bootstrap -s" in quiet mode.
​9c39bc8	Trac #30533: hide info output from "./bootstrap -D" in quiet mode.
​dd96aa3	Merge branch 't/30533/ticket/30533' into t/30664/fixup_for__add_quiet_mode_for_bootstrap_
​2278451	boostrap: Repair ./bootstrap in unconfigured source trees.

[Michael Orlitzky]

What error did you see with ./bootstrap? I tested it with every shell available to me, in a clean tree (after git clean -x -f -d).

Probably related: those two occurrences of source should be . instead. And we should be doing test -f <file> && . <file> instead of 2>/dev/null so that we don't hide other, unexpected errors.

[Matthias Köppe]

It just stops silently when trying to source the non-existing src/bin/sage-env-config.

$ ./bootstrap  -d
rm -rf config configure build/make/Makefile-auto.in
rm -f src/doc/en/installation/*.txt
rm -rf src/doc/en/reference/spkg/*.rst
rm -f src/doc/en/reference/repl/*.txt

[Michael Orlitzky]

I can't reproduce that with dash, bash, ksh, zsh, or posh. If you drop the 2>/dev/null, what goes wrong? (What shell are you using?)

The source is a bashism so I would think it fails there, but I don't know why it wouldn't proceed. In any case, changing those lines to -f <file> && . <file> will avoid the failure.

[Matthias Köppe]

I observed it both locally on macOS, and in tox -e docker-conda-forge-ubuntu-standard.

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​28e584c

bootstrap: Check for src/bin/sage-env-config before sourcing

[Michael Orlitzky]

I can't explain this, but when my /bin/sh points to /bin/zsh and I run the bootstrap script, it fails. If I run zsh ./bootstrap -d instead, it works fine -- so whatever I was doing to test was flawed. At first I thought zsh was looking at the shebang and re-executing the script with /bin/sh... but /bin/sh is still zsh!

Whatever is happening there, your branch fixes the problem. Look good on your end too?

[Matthias Köppe]

Often $0=sh enables posixisms. bash does the same.

[Matthias Köppe]

Yes, it works for me with this change, on both platforms

[Michael Orlitzky]

Ok, thanks. I'm still not sure why zsh handles that one set -e differently. I went back and re-tested the others, all as /bin/sh. More important things to worry about.

[Matthias Köppe]

Thanks!

[Dima Pasechnik]

Is there are make target to remove src/bin/sage-env-config without running ./boostrap ? I could not find one, and this is a problem - ./bootstrap can fail on a stale sage-env-config.

[Dima Pasechnik]

please see #30795 for a followup