mozjs52: update to 52.9.0.

[Cogitri]

No description provided.

[maxice8]

please use SPDX short-identifiers

[Cogitri]

This also enables jemalloc for glibc builds, since jemalloc's performance is still better than glibc's malloc. Should be tested once icu rebuilding has been done.

[maxice8]

You will need this patch, put it in the patches subdirectory with a .patch prefix.

--- memory/mozjemalloc/jemalloc.c	2018-09-07 01:30:40.056777558 -0300
+++ memory/mozjemalloc/jemalloc.c	2018-09-07 01:30:45.847777478 -0300
@@ -331,9 +331,6 @@
 #endif
 #include <sys/time.h>
 #include <sys/types.h>
-#if !defined(MOZ_MEMORY_SOLARIS) && !defined(MOZ_MEMORY_ANDROID)
-#include <sys/sysctl.h>
-#endif
 #include <sys/uio.h>
 #ifndef MOZ_MEMORY
 #include <sys/ktrace.h> /* Must come after several other sys/ includes. */

[Gottox]

[...] since jemalloc's performance is still better than glibc's malloc.

Are there any (real world) benchmarks that support that statement? The performance of memory allocation heavily depends on the workload. And even then I don't see where mozjs52 is used in performance critical packages.

Allows jemalloc to be used in conjunction with glibc malloc? What if mozjs52 allocates memory and the application uses the libc's free()? Does this work?

[Cogitri]

Are there any (real world) benchmarks that support that statement?

I can't seem to find benchmarks for spidermonkey with and without jemalloc right now, but other packages do have rather big performance improvements.

Also, Mozilla has been using jemalloc by default for years now, so I feel like it's a good idea if we use it as well.

I don't see where mozjs52 is used in performance critical packages.

Hm, yeah, once gjs 1.54.0 is in the tree the only packages consuming it seem to be cjs,goffie&polkit. If you don't want to pull in the dep for that that's fine by me

[maxice8]

I think it is better we just not enable jemalloc for now

[maxice8]

Pushed a fixed version of your commit:

• Removed patch that only apply to jeamalloc and was failing because patch-args is '-Np1' but the patch itself is meant for '-p0'