Linux: add gyp flag for enabling tcmalloc.

Linux: add gyp flag for enabling tcmalloc.  Also fix the linux build which didn't work in a clean client since tcmalloc.h wasn't generated.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=24538

[willchan no longer on Chromium, 2009-08-25 20:29:34]

I'm not sure if this is the right way to do this.  I considered trying to keep
all the conditional logic within tcmalloc.gyp and having base.gyp:base always
depend on tcmalloc.gyp:tcmalloc in linux, but tcmalloc.gyp:tcmalloc is sorta
fugly right now, with various windows and linux specific modifications, so I
didn't know of a clean way to do all the conditional logic in tcmalloc.gyp.  Let
me know if there's a better way to do this.

[willchan no longer on Chromium, 2009-08-25 20:54:45]

Oops, looks like this is broken due to some local files I have in my client that
make it work.  I need to add a tcmalloc.gyp fix.  This is still reviewable,
although I'll have to add mbelshe later for the tcmalloc.gyp change.

On 2009/08/25 20:29:34, willchan wrote:
> I'm not sure if this is the right way to do this.  I considered trying to keep
> all the conditional logic within tcmalloc.gyp and having base.gyp:base always
> depend on tcmalloc.gyp:tcmalloc in linux, but tcmalloc.gyp:tcmalloc is sorta
> fugly right now, with various windows and linux specific modifications, so I
> didn't know of a clean way to do all the conditional logic in tcmalloc.gyp. 
Let
> me know if there's a better way to do this.

[Mark Mentovai, 2009-08-25 21:03:08]

http://codereview.chromium.org/173387/diff/1/3
File build/all.gyp (right):

http://codereview.chromium.org/173387/diff/1/3#newcode55
Line 55: '../third_party/tcmalloc/tcmalloc.gyp:tcmalloc',
Why not :*?

[willchan no longer on Chromium, 2009-08-25 21:06:27]

http://codereview.chromium.org/173387/diff/1/3
File build/all.gyp (right):

http://codereview.chromium.org/173387/diff/1/3#newcode55
Line 55: '../third_party/tcmalloc/tcmalloc.gyp:tcmalloc',
On 2009/08/25 21:03:08, Mark Mentovai wrote:
> Why not :*?

The other target in tcmalloc.gyp is libcmt.  This is the windows runtime
library.  In windows, we strip out the malloc() family of functions from libcmt
(which is what prep_libc.sh does).  We don't need to do this in any other
platforms (glibc defines malloc() and friends as weak symbols).

[Mark Mentovai, 2009-08-25 21:21:05]

Why isn't the libcmt target inside an OS=3D=3D"win" conditional then?

On Tue, Aug 25, 2009 at 5:06 PM, <willchan@chromium.org> wrote:
>
> http://codereview.chromium.org/173387/diff/1/3
> File build/all.gyp (right):
>
> http://codereview.chromium.org/173387/diff/1/3#newcode55
> Line 55: '../third_party/tcmalloc/tcmalloc.gyp:tcmalloc',
> On 2009/08/25 21:03:08, Mark Mentovai wrote:
>>
>> Why not :*?
>
> The other target in tcmalloc.gyp is libcmt. =A0This is the windows runtim=
e
> library. =A0In windows, we strip out the malloc() family of functions fro=
m
> libcmt (which is what prep_libc.sh does). =A0We don't need to do this in
> any other platforms (glibc defines malloc() and friends as weak
> symbols).
>
> http://codereview.chromium.org/173387
>

[willchan no longer on Chromium, 2009-08-25 21:34:01]

mbelshe: Please review the tcmalloc.gyp and tcmalloc.h (generated from
tcmalloc.h.in by ./configure) changes.

On 2009/08/25 21:21:05, Mark Mentovai wrote:
> Why isn't the libcmt target inside an OS=3D=3D"win" conditional then?

Because I'm gyp illiterate and clearly have no idea about the best way to
accomplish things with gyp :)  I've changed all.gyp now so linux depends on
tcmalloc.gyp:* and made the libcmt target conditional on windows.  I haven't
been able to do another try job since the servers are down now, but I'll run
them later.

> 
> On Tue, Aug 25, 2009 at 5:06 PM, <mailto:willchan@chromium.org> wrote:
> >
> > http://codereview.chromium.org/173387/diff/1/3
> > File build/all.gyp (right):
> >
> > http://codereview.chromium.org/173387/diff/1/3#newcode55
> > Line 55: '../third_party/tcmalloc/tcmalloc.gyp:tcmalloc',
> > On 2009/08/25 21:03:08, Mark Mentovai wrote:
> >>
> >> Why not :*?
> >
> > The other target in tcmalloc.gyp is libcmt. =A0This is the windows runtim=
> e
> > library. =A0In windows, we strip out the malloc() family of functions fro=
> m
> > libcmt (which is what prep_libc.sh does). =A0We don't need to do this in
> > any other platforms (glibc defines malloc() and friends as weak
> > symbols).
> >
> > http://codereview.chromium.org/173387
> >

[Mark Mentovai, 2009-08-25 21:36:19]

LGTM if the answer is "yes."

http://codereview.chromium.org/173387/diff/8/10
File build/all.gyp (right):

http://codereview.chromium.org/173387/diff/8/10#newcode55
Line 55: '../third_party/tcmalloc/tcmalloc.gyp:*',
OK.  Is it safe to build tcmalloc when linux_use_tcmalloc is 0 in the other
file?

[willchan no longer on Chromium, 2009-08-26 20:04:18]

mbelshe: ping

On 2009/08/25 21:36:19, Mark Mentovai wrote:
> LGTM if the answer is "yes."
> 
> http://codereview.chromium.org/173387/diff/8/10
> File build/all.gyp (right):
> 
> http://codereview.chromium.org/173387/diff/8/10#newcode55
> Line 55: '../third_party/tcmalloc/tcmalloc.gyp:*',
> OK.  Is it safe to build tcmalloc when linux_use_tcmalloc is 0 in the other
> file?

[willchan no longer on Chromium, 2009-08-26 20:04:29]

http://codereview.chromium.org/173387/diff/8/10
File build/all.gyp (right):

http://codereview.chromium.org/173387/diff/8/10#newcode55
Line 55: '../third_party/tcmalloc/tcmalloc.gyp:*',
On 2009/08/25 21:36:19, Mark Mentovai wrote:
> OK.  Is it safe to build tcmalloc when linux_use_tcmalloc is 0 in the other
> file?

Yes.

[Mike Belshe, 2009-08-26 21:51:40]

lgtm