177486 - configure script should check for minimum Sun compiler version

Status: RESOLVED FIXED

(SeaMonkey :: Build Config, enhancement, P3)

Description

[Danek Duvall]

This is a followup to bug 177141.  During a compile, when configure runs, it
should check to make sure that the version of the Forte compilers being used is
at the minimum level (currently 6.2), and abort with an error message asking the
user to upgrade if the compilers are older.  This probably could be extended to
other compilers as well, but this is the one *I* ran into, so hey ... :)

Comment 1

[hacker formerly known as [e-mail]]

Attachment: v1.0

Comment 2

[Roland Mainz]

Comment on attachment 118884 [details] [diff] [review]
v1.0

Trying the patch with Sun Workshop 7 (official name is "Sun Forte 7") does not
work:
-- snip --
AWK=nawk ; CC_VERSION=`$CC -V 2>&1 | grep Sun 2>/dev/null | $AWK -F\: '{ print
$2 }'` ; CXX_VERSION=`$
CXX -V 2>&1 | grep Sun 2>/dev/null | $AWK -F\: '{ print $2 }'` ;
_MAJOR_VERSION_NUMBER=`echo $CC_VERSION | sed 's|^Sun WorkShop||' 
| $AWK '{ print $1 }'` ; /bin/echo
"CC_VERSION=$CC_VERSION\nCXX_VERSION=$CXX_VERSION\n_MAJOR_VERSION_NUMBER=$_MAJO
R_VERSION_NUMBER"

CC_VERSION=
CXX_VERSION=
_MAJOR_VERSION_NUMBER=
-- snip --

cc/CC version output for Sun Forte 7 looks like this:
-- snip --
% cc -V
cc: Forte Developer 7 C 5.4 2002/03/09
usage: cc [ options] files.  Use 'cc -flags' for details
% CC -V
CC: Forte Developer 7 C++ 5.4 Patch 111715-05 2003/02/09
-- snip --

Comment 3

[hacker formerly known as [e-mail]]

Did I ever mention how Sun's compiler team needs a boot to the head?  Irregular
version output and they change the name between releases.  Joy oh joy.

Comment 4

[hacker formerly known as [e-mail]]

Attachment: v1.1

Comment 5

[Roland Mainz]

Christopher Seawood wrote:
> Did I ever mention how Sun's compiler team needs a boot to the head?  
> Irregular version output and they change the name between releases. 

Please beat Sun marketing droids first. They renamed "Sun Workshop" to "Sun
Forte" and then to "Sun ONE" ; poor compiler folks at Sun just has to implement
their ******** inventions... ;-(

Comment 6

[Roland Mainz]

Comment on attachment 119154 [details] [diff] [review]
v1.1

r=[e-mail]
The patch should work with Workshop/Forte 7 (and hopefully with Workshop/Forte
8 and >= 6.2, too, but I don't have these compilers installed and can't test
that here...) ...

Comment 7

[hacker formerly known as [e-mail]]

Checked in.

Comment 8

[Greg Onufer]

The trunk has been broken since this was checked in.  First, the carets aren't
quoted so configure fails (see the nebiros logs):

diff -u -3 -p -u -r1.1207 configure.in
--- configure.in        4 Apr 2003 21:14:34 -0000       1.1207
+++ configure.in        5 Apr 2003 03:08:22 -0000
@@ -1394,8 +1394,8 @@ dnl the qsort routine under solaris is f
        AS='/usr/ccs/bin/as'
        ASFLAGS="$ASFLAGS -K PIC -L -P -D_ASM -D__STDC__=0"
        AS_DASH_C_FLAG=''
-        CC_VERSION=`$CC -V 2>&1 | grep ^cc: 2>/dev/null | $AWK -F\: '{ print $2
 }'`
-        CXX_VERSION=`$CXX -V 2>&1 | grep ^CC: 2>/dev/null | $AWK -F\: '{ print 
$2 }'`
+        CC_VERSION=`$CC -V 2>&1 | grep '^cc:' 2>/dev/null | $AWK -F\: '{ print 
$2 }'`
+        CXX_VERSION=`$CXX -V 2>&1 | grep '^CC:' 2>/dev/null | $AWK -F\: '{ prin
t $2 }'`
         _MAJOR_VERSION_NUMBER=`echo $CC_VERSION | $AWK '{ print $3 }'`
         _BAD_COMPILER=

Next, the version fields aren't stable and this fails with the next Workshop
where it gets "C" (language name) instead of the compiler version.

Comment 9

[Roland Mainz]

Greg Onufer wrote:
> The trunk has been broken since this was checked in.  First, the carets aren't
> quoted so configure fails (see the nebiros logs):

Ahhhhgrrlll... my fault... ;-(

> diff -u -3 -p -u -r1.1207 configure.in
> --- configure.in        4 Apr 2003 21:14:34 -0000       1.1207
> +++ configure.in        5 Apr 2003 03:08:22 -0000
> @@ -1394,8 +1394,8 @@ dnl the qsort routine under solaris is f
>         AS='/usr/ccs/bin/as'
>         ASFLAGS="$ASFLAGS -K PIC -L -P -D_ASM -D__STDC__=0"
>         AS_DASH_C_FLAG=''
> -        CC_VERSION=`$CC -V 2>&1 | grep ^cc: 2>/dev/null | $AWK -F\: '{ print
> $2
>  }'`
> -        CXX_VERSION=`$CXX -V 2>&1 | grep ^CC: 2>/dev/null | $AWK -F\: '{ 
> print $2 }'`
> +        CC_VERSION=`$CC -V 2>&1 | grep '^cc:' 2>/dev/null | $AWK -F\: '{ 
> print  $2 }'`
> +        CXX_VERSION=`$CXX -V 2>&1 | grep '^CC:' 2>/dev/null | $AWK -F\: '{ 
> print $2 }'`
>          _MAJOR_VERSION_NUMBER=`echo $CC_VERSION | $AWK '{ print $3 }'`
>          _BAD_COMPILER=

Sounds reasonable...

> Next, the version fields aren't stable and this fails with the next Workshop
> where it gets "C" (language name) instead of the compiler version.

What do you suggest what we should do instead ? Anything lower than Sun Workshop
6 Update 2 can't compile Mozilla properly and if we do not add such a check the
people won't stop filing bugs for their broken compilers... ;-(

Comment 10

[Harshal Pradhan]

Attachment: bustage fix

For the record I checked this in, in an attempt to fix nebiros bustage. Its
lets my tree build with F6U2 again and should hopefully clear up the red for
now.

Comment 11

[Roland Mainz]

Harshal Pradhan wrote:
> For the record I checked this in, in an attempt to fix nebiros bustage. Its
> lets my tree build with F6U2 again and should hopefully clear up the red for
> now.

Thanks for rescuing me... :)

Comment 12

[Roland Mainz]

Alternate solution for the compiler version check may be to create a small
application, compile it with the compiler and let it check the compiler version
that way. The quiz is which values are used for __SUNPRO_CC in Sun Workshop 6
Update 2 and lower... ;-/

Comment 13

[Danek Duvall]

The actual token is __SUNPRO_C (not CC), and here are some test values:

  Sun Workshop 4.2:  0x420
  Sun Workshop 5.0:  0x500
  Sun Workshop 6.1:  0x520
  Sun Workshop 6.2:  0x530
  Forte Developer 7: 0x540
  SunONE Studio 8:   0x550

Also looks like the output of cc -V for all these compiler versions include the
string "C <version>" in the first line of output, where 5.3 corresponds to
Workshop 6 update 2.  Not sure which method is easier, but at least __SUNPRO_C
is a documented macro, and is more likely to remain useful across compiler
versions than the obviously ever-changing version output.

Comment 14

[Danek Duvall]

Um, that is, if you use the C compiler.  __SUNPRO_CC *is* the right token for
C++, and it produces the same numbers as __SUNPRO_C does for the C compiler.

Comment 15

[Kyle Yuan]

this version checking is totally wrong. 

FYI: my cc -V said
cc: Sun WorkShop 6 update 2 C 5.3 2001/05/15
"Sun WorkShop 6 update 2" is the compiler name, not the version, "5.3" is.

For the next version compiler (internal called K2), cc -V output:
cc: Sun C 5.5 2003/03/09
"5.5" is the version number. (same thing with the CC -V)

So this checkin will break the build for Sun's new compiler.

BTW, __SUNPRO_C/__SUNPRO_CC does work for both F6U2 & K2.

Comment 16

[hacker formerly known as [e-mail]]

Attachment: v2.0

Untested patch to change the version test to use __SUNPRO_C.

Comment 17

[Kyle Yuan]

Comment on attachment 122773 [details] [diff] [review]
v2.0

cls, the configure doesn't work for me. I don't know why, but the following
piece of code does work with both WS6U2 & K2. (I copied them from the configure
you sent to me)

int main() {
#if (__SUNPRO_C < 0x530)
	   #error "Denied"
	   #endif
; return 0; 
}

Comment 18

[Roland Mainz]

Comment on attachment 122773 [details] [diff] [review]
v2.0

review- per  Kyle Yuan's comment #17 ... I have to check that in detail (each
time I touch the build system tinderboxen  get a red angry face, so I'd like to
be carefully now... ;-/ )

Comment 19

[hacker formerly known as [e-mail]]

Roland, the review request was so that you could check the patch in detail (ie,
test it).  Kyle can't run autoconf 2.13 and I have no idea why the generated
configure I sent him doesn't work on his box.

Comment 20

[Kyle Yuan]

sorry, I was in a hurry to test this patch before. Now I'd like to find out why
it doesn't work.

Comment 21

[Kyle Yuan]

Comment on attachment 122773 [details] [diff] [review]
v2.0

I've installed autoconf 2.13 on my solaris box and found out that some missing
brackets break this patch. After I corrected that, it works with both WS6U2 &
K2.

r=kyle with below changed:

>+       AC_MSG_CHECKING([for Forte compiler version >= WS6U2]
missed a ')' at the end of line

>+        AC_MSG_RESULT[$_res])
missed a '(' before |[$res])|

Comment 22

[hacker formerly known as [e-mail]]

Kyle, thanks for catching my oversights.  The modified patch has been checked in.

Comment 23

[cls]

*** Bug 212199 has been marked as a duplicate of this bug. ***

Comment 24

[cls]

*** Bug 216562 has been marked as a duplicate of this bug. ***

Comment 25

[Kyle Yuan]

Comment on attachment 122773 [details] [diff] [review]
v2.0

nominate for 1.4.x.

Comment 26

[Asa Dotzler [:asa]]

Comment on attachment 122773 [details] [diff] [review]
v2.0

This is not going to make 1.4.1.  Please re-request aproval after 1.4.1 ships
if you'd like to get this in for 1.4.2.