[Bug mi/29426] FAIL: gdb.mi/mi-var-invalidate.exp: linteger not anymore in scope due to binary changes (unexpected output)

["lsix at lancelotsix dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29426

--- Comment #8 from Lancelot SIX <lsix at lancelotsix dot com> ---
Created attachment 14242
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14242&action=edit
add_varobj_global_flag.patch

This regression also shows an interesting problem.

Currently, when we unload an objfile and find a varobj which is linked to a
block from this objfile, we clear the valid_block pointer to avoid having a
stall pointer.  However by doing this, we loose the ability to distinguish a
varobj tracking a global (no valid_block associated) to a varobj tracking a
local (some valid_block associated).  Later, when we reload a new program and
try to recreate varobj for globals, we have false positive and also try to
recreate some locals as well.

The attached patch fixes this, but some differences remain whether I ubuntu’s
stock GCC or custom built HEAD GCC.

-- 
You are receiving this mail because:
You are on the CC list for the bug.