On Fri, 9 May 2025, Pali Rohár wrote:
On Tuesday 06 May 2025 19:58:15 Martin Storsjö wrote:
On Thu, 17 Apr 2025, Pali Rohár wrote:
This allows to remove the whole pre_c_init() and pre_cpp_init() functions
and registration of __mingw_pcinit and __mingw_pcppinit hooks.
Inlining the code allows to easily figure out in which order is
initialization code called.
Also merge code which is protected by __native_startup_lock into one
if-branch for better code readability.
Function pre_c_init() was the first callback in .CRT$XI section, so now
call its code before the _initterm_e(__xi_a, __xi_z) (which executes all
remaining .CRT$XI callbacks).
Function pre_cpp_init() was the first callback in .CRT$XC section, so now
call its code before the _initterm (__xc_a, __xc_z) (which executes all
remaining .CRT$XC callbacks).
---
mingw-w64-crt/crt/crtexe.c | 98 +++++++++++++++-----------------------
1 file changed, 38 insertions(+), 60 deletions(-)
I'm not fully convinced about this one; we could also leave the pre_c_init
and pre_cpp_init functions as they were, but just call them directly instead
of registering them in the .CRT$.. sections; that makes it clearer exactly
where/when they are called - without moving code around without any apparent
purpose. I guess it depends on the overall final direction of this patchset
if that makes things any better in any tangible way?
// Martin
I was thinking about it. But somehow for me it was cleaner and easier to
understand order of the executed functions after "inlining" that code.
That was just my opinion about this. Also same thing I have already done
for crtdll.c startup code and those changes were already merged.
I'm still targeting to fix the security issue related to cmdline
processing and for this work I first wanted to do cleanups which allows
me to better understand the code and sequence when which thing happens
and so what is the correct stage to fix the cmdline processing.
Fair enough, I guess it's ok to inline it like this, for that final target
of making the end result easier to reason about (which I do agree that it
does).
// Martin
_______________________________________________
Mingw-w64-public mailing list
Mingw-w64-public@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
