Re: Exit status ignored during variable expansion?

[Angel Tsankov]

On 07/28/11 22:28, Paul Smith wrote:
> On Thu, 2011-07-28 at 22:18 +0300, Angel Tsankov wrote:
> > Does anyone have an idea why 'make' ignores the exit status from any
> > commands that are run as part of variable expansion:
> >
> > a=$(shell no-such-command)
> > b=$(shell false)
>
> What do you think make should do with the exit status of these
> operations?  These are not run as part of any rule or target.  I suppose
> if it had been documented/worked since the beginning that non-0 exit
> codes from the shell function would cause make to fail, that would have
> been OK.
>
> But I don't see any clear enough advantage to behaving that way to
> justify a massively backward-incompatible change like that.


If 'make' does not fail on non-zero exit from a command (during variable 
expansion) then we have no way to determine if the command completed 
successfully (with empty output) or if it failed.  This is my only concern.

A possible solution would be to introduce a new target (such as 
.LOW_RESOLUTION_TIME) for this behavior.


Regards,
Angel Tsankov