On 10/24/2017 05:57 PM, Emmanuel Hocdet wrote: > >> Le 3 août 2017 à 10:07, Willy Tarreau <w...@1wt.eu <mailto:w...@1wt.eu>> a >> écrit : >> >> Hi Bernard, >> >> I'm CCing Emeric since this affects SSL. I have some comments below. >> >> On Tue, Jul 25, 2017 at 05:03:10PM +0200, Bernard Spil wrote: >> >>> --- src/ssl_sock.c.orig2017-06-02 13:59:51 UTC >>> +++ src/ssl_sock.c >>> @@ -56,7 +56,7 @@ >>> #include <openssl/engine.h> >>> #endif >>> >>> -#if OPENSSL_VERSION_NUMBER >= 0x1010000fL >>> +#if (OPENSSL_VERSION_NUMBER >= 0x1010000fL) && >>> !defined(LIBRESSL_VERSION_NUMBER) >>> #include <openssl/async.h> >>> #endif >> (...) >> >> OK so all the ones related to LibreSSL not supporting async should go >> into one patch. > > > "[PATCH] MINOR: ssl: build with recent BoringSSL library » used > OPENSSL_NO_ASYNC > > With this patch, a generic way could be to have (if LibreSSL have no > OPENSSL_NO_ASYNC defined): > > #ifdef LIBRESSL_VERSION_NUMBER > > #ifndef OPENSSL_NO_ASYNC > #define OPENSSL_NO_ASYNC > #endif > > #endif > > > I think you wanna mean:
#if (OPENSSL_VERSION_NUMBER < 0x1010000fL || defined LIBRESSL_VERSION_NUMBER) #ifndef OPENSSL_NO_ASYNC #define OPENSSL_NO_ASYNC #endif #endif and review all subtests to check OPENSSL_NO_ASYNC. I think it is a good idea. R, Emeric