shithub: libvpx

Info  •  Files  •  Log  •  Branches

Fix windows build: obj_int_extract to extract numbrs from .rdata

The change in assembly offset files to define values as const int broke
Windows build, because the variables are stored in .rdata section instead
of .data section.

This CL changes the integer peeking from .data to .rdata.

Change-Id: I87e465ddcc78d39ec29f3720ea7df0ab807d5512

--- a/build/make/obj_int_extract.c

+++ b/build/make/obj_int_extract.c

@@ -680,7 +680,7 @@

     uint32_t symoffset;

     char **sectionlist;  //this array holds all section names in their correct order.

-    //it is used to check if the symbol is in .bss or .data section.

+    //it is used to check if the symbol is in .bss or .rdata section.

     nsections = get_le16(buf + 2);

     symtab_ptr = get_le32(buf + 8);

@@ -725,15 +725,15 @@

         }

         strcpy(sectionlist[i], sectionname);

-        if (!strcmp(sectionname, ".data")) sectionrawdata_ptr = get_le32(ptr + 20);

+        if (!strcmp(sectionname, ".rdata")) sectionrawdata_ptr = get_le32(ptr + 20);

         ptr += 40;

     }

     //log_msg("COFF: Symbol table at offset %u\n", symtab_ptr);

-    //log_msg("COFF: raw data pointer ofset for section .data is %u\n", sectionrawdata_ptr);

+    //log_msg("COFF: raw data pointer ofset for section .rdata is %u\n", sectionrawdata_ptr);

-    /*  The compiler puts the data with non-zero offset in .data section, but puts the data with

+    /*  The compiler puts the data with non-zero offset in .rdata section, but puts the data with

         zero offset in .bss section. So, if the data in in .bss section, set offset=0.

         Note from Wiki: In an object module compiled from C, the bss section contains

         the local variables (but not functions) that were declared with the static keyword,

-- 

⑨