[Lldb-commits] [lldb] [LLDB] Impove ObjectFileELF's .dynamic parsing and usage. (PR #101237)

[Pavel Labath via lldb-commits]

================
@@ -2550,6 +2550,21 @@ ModuleSP Process::ReadModuleFromMemory(const FileSpec &file_spec,
   }
   ModuleSP module_sp(new Module(file_spec, ArchSpec()));
   if (module_sp) {
+    if (size_to_read == 0) {
+      // Default to 8192 in case we can't find a memory region.
+      size_to_read = 0x2000;
+      MemoryRegionInfo range_info;
+      Status error(GetMemoryRegionInfo(header_addr, range_info));
+      if (error.Success()) {
+        // We found a memory region, set the range of bytes ro read to read to
+        // the end of the memory region. This should be enough to contain the
+        // file header and important bits.
+        const auto &range = range_info.GetRange();
----------------
labath wrote:

Depending on how the file is linked (`-z noseparate-code`) the first region could end up containing all of the code in the binary (and so be arbitrarily big), it might not be a good idea to read all of it. What's the reason for this change? Could that be a separate patch as well? (I think we can test DT_DYNAMIC parsing without actually loading the file from memory)

For reference, this is what I get with a hello world binary built like this:
```
$ clang++ -o /tmp/a.out -x c++ - -static <<<"int main(){}" -Wl,-z,noseparate-code && llvm-readelf -e /tmp/a.out  | grep -e LOAD -B 1
  Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
  LOAD           0x000000 0x0000000000400000 0x0000000000400000 0x09caf6 0x09caf6 R E 0x1000
  LOAD           0x09ceb0 0x000000000049deb0 0x000000000049deb0 0x005b58 0x00b2b8 RW  0x1000
```

https://github.com/llvm/llvm-project/pull/101237