Tom Lane writes:
> Also, I'm still feeling that we are missing something fundamental about
> parallel make. Surely there's got to be *some* interlock in make that
> prevents multiple subjobs from executing the same rule in parallel.
There is, but remember that the two-target rule we're looking at is
actually two separate rules written in abbreviated form. And under that
pretext, the two rules are lying: They don't generate exactly the file
they have as the target. Unfortunately, there isn't a right way to fix
Keep in mind that parallel make only executes the rule commands in
parallel, it does not cause the dependency analysis to be distributed.
In order to satisfy a dependency graph you only need to process each node
once, and parallelism doesn't change that.
Peter Eisentraut peter_e(at)gmx(dot)net
In response to
pgsql-patches by date
|Next:||From: Bruce Momjian||Date: 2001-11-12 19:23:23|
|Subject: Re: [ODBC] MD5 support for ODBC|
|Previous:||From: Peter Eisentraut||Date: 2001-11-12 19:22:13|
|Subject: Re: Patch for Makefile race against current cvs |