> -----Original Message----- > From: Laszlo Ersek [mailto:[email protected]] > Sent: Thursday, February 25, 2016 2:01 AM > To: David Woodhouse <[email protected]>; edk2-devel-01 > <[email protected]> > Cc: Eric Dong <[email protected]>; Cecil Sheng <[email protected]>; Ting > Ye > <[email protected]>; Qiu Shumin <[email protected]>; Qin Long > <[email protected]>; Liming Gao <[email protected]>; Yao Jiewen > <[email protected]>; Daryl McDaniel <[email protected]>; Jaben > Carsey > <[email protected]>; Samer El-Haj-Mahmoud <[email protected]> > Subject: Re: [edk2] [PATCH 0/4] free(NULL) and realloc(NULL, size) conformance > improvements > > On 02/25/16 03:04, David Woodhouse wrote: > > On Wed, 2016-02-24 at 22:13 +0100, Laszlo Ersek wrote: > >> The free() wrapper in BaseCryptLib has a bug that has been triggered > >> by David's recent OpenSSL work. The series fixes the bug, plus > >> more instances of the same. > > > > Should we not just fix the underlying FreePool() function to do the > > sane thing. > > It crossed my mind, but FreePool() is extremely widely used, it has a > detailed interface contract, and its current behavior matches its > interface contract. I'm not up to auditing all uses of FreePool(), to > find the one that perversely enforces a failure return form > FreePool(NULL). :) > > > Anyway, I've rebased my tree on top of yours, > > Thanks -- I'll push the first three patches to edk2 master in a minute, > and I'll post a new version of the fourth. > > > split up the patch > > changes into separate bisectable commits, and pushed my tree out again > > to http://git.infradead.org/users/dwmw2/edk2.git > > I skimmed your fresh master -- the way the patch evolves looks > excellent. I guess it took a lot of effort. > > > Both Cryptest.efi and the test boot of Fedora 22 are working correctly > > at all stages. > > Perfect. > > > Again, the final two commits aren't ready yet. But the rest probably > > are if they build OK on Windows. > > > > I do still want to kill that -w. And why in $DEITY's name do we not > > already have -nostdinc in our CFLAGS for the whole EDK2 build?
The -nostdinc equivalent for Visual Studio breaks the NT32 build and there is no way to negate the flag once it has been specified. There would have to be a new target added just for NT32 builds. > I propose to involve our BaseTools overlords here... > > Thanks > Laszlo Daryl McDaniel _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
