When building a shared ELF library with FIPS, set the section markers hidden.

This forces resolving of these marker symbols within the BCM to ignore
other libraries, thereby making the hash computation correct even if
multiple BoringSSL builds were to be linked together.

Bug: 42220000
Change-Id: I02cd6941bb0b894492ae98d00c1727fa6a6a6964
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/89227
Commit-Queue: Rudolf Polzer <rpolzer@google.com>
Reviewed-by: David Benjamin <davidben@google.com>

crypto/fipsmodule/fips_shared.lds

diff --git a/crypto/fipsmodule/fips_shared.lds b/crypto/fipsmodule/fips_shared.lds
index 1f20b96..4f9c0a1 100644
--- a/crypto/fipsmodule/fips_shared.lds
+++ b/crypto/fipsmodule/fips_shared.lds
@@ -1,20 +1,20 @@
 SECTIONS
 {
   .text : {
-    BORINGSSL_bcm_text_start = .;
+    PROVIDE_HIDDEN(BORINGSSL_bcm_text_start = .);
     *(.text)
     *(.text.unlikely.*)
     /* These sections shouldn't exist. But C++ `inline` symbols seem to be
      * placed in function sections on Android even with
      * -fno-function-sections. */
     *(text.*)
-    BORINGSSL_bcm_text_end = .;
+    PROVIDE_HIDDEN(BORINGSSL_bcm_text_end = .);
   }
   .rodata : {
-    BORINGSSL_bcm_rodata_start = .;
+    PROVIDE_HIDDEN(BORINGSSL_bcm_rodata_start = .);
     *(.rodata)
     *(.rodata.*)
-    BORINGSSL_bcm_rodata_end = .;
+    PROVIDE_HIDDEN(BORINGSSL_bcm_rodata_end = .);
   }
 
   /DISCARD/ : {