112244 - Update config.guess & config.sub to standard

Status: VERIFIED FIXED

(SeaMonkey :: Build Config, defect)

Description

[jhp (no longer active)]

At the moment, config.guess returns a string that looks like "i386-pc-os2_emx"
or "i386-pc-os2_vacpp".  A while back, while talking to one of the config.*
developers, I was informed that it is not standard to include any refernce to
the compiler in the string.  This needs to be changed to match other platforms.

Comment 1

[jhp (no longer active)]

Attachment: patch

I changed to using the generic string "i386-pc-os2" and determining the
difference between an EMX/GCC and VisualAge C++ build by testing the $CC
variable.  I am also going to submit the config.* patches to the maintainers to
get their input.

Comment 2

[hacker formerly known as [e-mail]]

Why not test against GNU_CC rather than explicitly checking for icc?  Your tests
will break if anyone tries to pass CC=/full/path/icc to configure.

Comment 3

[jhp (no longer active)]

GNU_CC seems to only be set if the environment variable GCC is set.  While
EMX/GCC may set this, it doesn't help with compilers that are not gcc.  In our
case, I decided to look at the value of CC.  You are right, though, that this
may be a full path.  Should we require developers to set an environment variable
(such as VACPP = yes) and then check for that in configure.in?

Comment 4

[jhp (no longer active)]

Attachment: another patch

OK, in this version of configure.in, I check for $GNU_CC and $VACPP.  A user of
EMX/GCC would set "$GCC" = "yes", while a user of Visual Age would set "$VACPP"
= "yes".  Also, I have taken the OS/2 specific changes from the current
versions of config.guess and config.sub.  How does this look?

Comment 5

[hacker formerly known as [e-mail]]

Testing for "$GNU_CC" should work when looking for non-gcc compilers as well
unless you're supporting multiple non-gcc compilers.  I didn't think that was
the case here.  I'd avoid making people explicitly set VACPP if possible.  

Comment 6

[jhp (no longer active)]

Well, I thought that assuming that any non-gcc compiler is VACPP would be wrong;
 that it would be best to explicitly check for the compiler.  Secondly, we
already set VACPP=yes.  In fact, what we do have to change is to set GCC=yes for
EMX/GCC.

Comment 7

[hacker formerly known as [e-mail]]

Comment on attachment 59415 [details] [diff] [review]
another patch

Ok, if you already require VACPP to be set, then the patch is fine as is. r=cls
Btw, configure automatically sets GCC=yes if it detects that $CC is a gcc
variant so there's no need to tell the developer to do so.

Comment 8

[hacker formerly known as [e-mail]]

Patch has been checked into the mozilla trunk, NSPR trunk and NSPR
NSPRPUB_PRE_4_2_CLIENT_BRANCH branch.

Comment 9

[jhp (no longer active)]

verified