----- Original Message -----
> From: [email protected]
> To: "Mike Frysinger" <[email protected]>
> Cc: [email protected]
> Sent: Tuesday, 23 April, 2013 8:08:43 PM
> Subject: Re: [LTP] [PATCH] remove linux_syscall_numbers.h from tree
> 
> Hi!
> > > > is keeping it in the tree a big deal ?  i think it's pretty common for
> > > > people to go into specific subdirs and run `make` and today, our
> > > > subdirs don't declare a dependency on this file so it'll get
> > > > automatically regenerated (nor do i think they should even bother).
> > > 
> > > You need to regenerate the file after each change to the *.in files,
> > > and that means that people will forget from time to time.
> > 
> > so people aren't testing their commits ;)
> 
> Well, I do not like to increase number of things to check for a commit.
> Fixing the testcases is complicated enough...
> 
> > > It would be better to add another rule to testcases.mk similiar to the
> > > one for LIBLTP.
> > 
> > those rules only generate libltp.a if the file doesn't already exist.  so
> > tweaking that would be like the status quo, but worse in that `git pull` is
> > now guaranteed to not update the file :).
> 
> Hmm, I would expect that the library will be rebuild after some of it's
> sources has changed. And we aren't doing that, I would call such
> behavior a bug.
> 
> But thinking of consequencies, changing it carelessly would most likely
> break parallel make.
> 
> > i don't want to add things like `make -C testcases/kernel/include/` as then
> > you'll have multiple subdirs processing the same thing.  maybe if we had a
> > .mk
> > file in testcases/kernel/include/ that testcases.mk could include, and if
> > the
> > top level testcases/kernel/ make sure we process that first before any
> > subdirs.
> 
> I see.
> 
> Or we get this behavior if we set a flag in toplevel Makefile and the
> header would be regenerated (in the leaf Makefile) only if the toplevel
> flag is not set. And we could do the same for the library as well.

One more issue I've came across is that, even when I had libltp.a and LSN_H
being rebuild just once from any directory when their sources are changed,
testcases didn't get rebuild too, because these are order-only prerequisites:
$(MAKE_TARGETS): | $(MAKE_DEPS)

Changing it to normal opens another can of worms - breaks implicit rules,
which are trying to e.g. link binary with datafiles/bin.sm or LSN_H.

Regards,
Jan

> 
> --
> Cyril Hrubis
> [email protected]
> 
> ------------------------------------------------------------------------------
> Try New Relic Now & We'll Send You this Cool Shirt
> New Relic is the only SaaS-based application performance monitoring service
> that delivers powerful full stack analytics. Optimize and monitor your
> browser, app, & servers with just a few lines of code. Try New Relic
> and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
> _______________________________________________
> Ltp-list mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/ltp-list
> 

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Reply via email to