make check discussion
jhuntwork at linuxfromscratch.org
Thu Jun 17 19:34:49 PDT 2004
On Thursday 17 June 2004 10:17 pm, Ryan.Oliver at pha.com.au wrote:
> > I'm sorry, but I don't understand.
> > "Normally we would now run the test suite, but at this early stage
> > the test suite framework (Tcl, Expect and DejaGNU) is not yet in
> > place. And there would be little point in running the tests anyhow,
> > since the programs from this first pass will soon be replaced by
> > those from the second. "
> Glibc doesn't use TCL/expect/dejagnu, though it does use perl for
> parts of the testsuite.
> > Why would this need to be changed? Is that statement wrong?
> Not for gcc and binutils (though really if you are gonna be doing
> any toolchain building your host SHOULD have the gnu test framework
> in place and you SHOULD check your preliminary toolchain components,
> especially if you are delving into "unstable" territory) but the
> statement _is_ wrong for glibc.
> As for the statement about there being little point checking them as
> they are going to be replaced.
> Fair enough gcc is pretty tolerant, as long as it can complete the
> build of xgcc during bootstrap and therefore compile itself from
> itself, last time I checked we don't do a bootstrap build of shared
> gcc in ch5.
> You can get away without a bootstrap for the shared build because we
> assume our ch5 static gcc is good, but note without testing it this
> is an assumption only.
> I know I'm being anal retentive here (hey, I'm known for it) but I
> don't like leaving anything to chance or base things on unproved
> assumptions wherever it can be avoided... thats why the build method
> is so robust, during its development absolutely everthing was tested
> and absolutely nothing left to chance.
So how would you have it be? Recommendations to test in both chapter 5
& 6, or optional in 5 and recommended in 6?
More information about the lfs-dev