https://bugs.kde.org/show_bug.cgi?id=514206

--- Comment #8 from Paul Floyd <[email protected]> ---
For the record the change that caused this to go from a warning to an error
dates back to the change that I made in
7844752299b5472b21fc4df765d4cffdf92c6c3d for split RW PT_LOAD segements. Prior
to that Valgrind assumed that exes have one RX and one RW PT_LOAD. Then some
linkers (lld, mold) started splitting the RW segment (I believe that this is
for RELRO support, the first contains dynamic fixups and it gets mapped RW then
when the fixups have been resolved it switches to RO, the second is the classic
.data and .bss).

With just 1 RW_PTLOAD we were just checking that the file being mapped is an
ELF one. Now we have to do some ELF parsing to get the number of expected RW
PT_LOADS.

The function that reads chunks of the mmap'd file from an fd has no failure
more. It either succeeds or it asserts. Previously, if we couldn't read the
start of the mmap'd file we would exit with a warning straight away.

Going back to just printing a warning would mean adding a failure mode to at
least set_CEnt() plus adding the code to handle failures in all callers.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to