master 83a6224: make-dist: check exit statuses more carefully

[Robert Pluim]

branch: master
commit 83a6224d607c645cadbe371c921928166da0aef0
make-dist: check exit statuses more carefully
* make-dist: Do a better job checking for subprocess failure.

Would 'set -e' not be easier?

Robert

[Glenn Morris]

branch: master
commit 83a6224d607c645cadbe371c921928166da0aef0

[...]

make-dist: check exit statuses more carefully

This breaks eg ./make-dist --no-update --tar --no-check.
Ref eg https://hydra.nixos.org/build/72642760

(There's no need for ChangeLog to exist in that usage.)

[Paul Eggert]

Post by Glenn Morris
This breaks eg ./make-dist --no-update --tar --no-check.
Ref eghttps://hydra.nixos.org/build/72642760

Thanks for mentioning that. I installed the attached, which should fix this.

[Paul Eggert]

Post by Robert Pluim
Would 'set -e' not be easier?

I've never trusted 'set -e' for anything other than simpleminded
debugging, because set -e is a global setting (so it's inflexible) and
it doesn't always work (so it's unreliable). For an example of the latter:

set -e; { false; true; } && echo x

This succeeds and outputs "x" even though the "false" command failed.
(POSIX requires this and it's longstanding behavior.)

[Robert Pluim]

Post by Paul Eggert

Post by Robert Pluim
Would 'set -e' not be easier?

I've never trusted 'set -e' for anything other than simpleminded
debugging, because set -e is a global setting (so it's inflexible) and
set -e; { false; true; } && echo x
This succeeds and outputs "x" even though the "false" command
failed. (POSIX requires this and it's longstanding behavior.)

Youʼre right. POSIX isnʼt even ambiguous about it.

Robert