lib: sync guard for Curl_getaddrinfo_ex() definition and use #9734
Conversation
|
Ref: #9727 (comment) |
vszakats
force-pushed
the
getaddrinfo-confusion
branch
from
ec960ae to
7f48f82
Compare
vszakats
added
the
feature-window
vszakats
force-pushed
the
getaddrinfo-confusion
branch
from
c3dfe76 to
ededf79
Compare
`Curl_getaddrinfo_ex()` gets _defined_ with `HAVE_GETADDRINFO` set. But, `hostip4.c` _used_ it with `HAVE_GETADDRINFO_THREADSAFE` set alone. It meant a build with the latter, but without the former flag could result in calling this function but not defining it, and failing to link. Patch this by adding an extra check for `HAVE_GETATTRINFO` around the call. Before this patch, build systems prevented this condition. Now they don't need to. Follow-up to 67d8862 Closes #xxxx
vszakats
force-pushed
the
getaddrinfo-confusion
branch
from
ededf79 to
ebded4c
Compare
| @@ -1079,9 +1079,6 @@ check_symbol_exists(_strtoi64 "${CURL_INCLUDES}" HAVE__STRTOI64) | |||
| check_symbol_exists(strerror_r "${CURL_INCLUDES}" HAVE_STRERROR_R) | |||
| check_symbol_exists(siginterrupt "${CURL_INCLUDES}" HAVE_SIGINTERRUPT) | |||
| check_symbol_exists(getaddrinfo "${CURL_INCLUDES}" HAVE_GETADDRINFO) | |||
| if(NOT HAVE_GETADDRINFO) | |||
| set(HAVE_GETADDRINFO_THREADSAFE OFF) | |||
| endif() | |||
There was a problem hiding this comment.
wouldn't it make sense to leave this
There was a problem hiding this comment.
It wouldn't hurt, but why guard against this twice?
Also without it, our CI wouldn't be really verifying this PR.
There was a problem hiding this comment.
I don't see where else it is checked. What I am saying is it wouldn't make sense that HAVE_GETADDRINFO_THREADSAFE is defined when not HAVE_GETADDRINFO.
There was a problem hiding this comment.
The root problem was that hostip4.c assumed that if HAVE_GETADDRINFO_THREADSAFE was set, also HAVE_GETADDRINFO was set. This was not always true, so I added this condition to CMake. (autotools was setting them in pair already). This still didn't prevent a manual/accidental misconfiguration. Hence this PR, where I address the root cause in hostip4.c, making sure to check that both are set, there. This also makes the manual CMake condition redundant.
[ You may ask, how could CMake set HAVE_GETADDRINFO_THREADSAFE if there was no getaddrinfo() present. The reason for that is that HAVE_GETADDRINFO_THREADSAFE is force-set to 1 on Windows, while the latter is properly auto-detected! ]
There was a problem hiding this comment.
Ok. IMO it wouldn't hurt to leave it there.
There was a problem hiding this comment.
Fine with me, I'll put it back there.
There was a problem hiding this comment.
@jay: Rephrasing the Windows logic as this:
if(WIN32)
set(HAVE_GETADDRINFO_THREADSAFE ${HAVE_GETADDRINFO})
endif()would IMO express our intent much cleaner and self contained in these three lines, with identical results. It would also allow dropping the related line from WindowsCache.cmake.
Meaning: When we have getaddrinfo(), it is always threadsafe (on Windows).
What do you think?
Simplify 67d8862, by setting `HAVE_GETADDRINFO_THREADSAFE` to the detection result of `HAVE_GETADDRINFO` on Windows. This expresses this intent clearer than the previous patch and keeps this logic in a single block of code: On Windows when we have `getaddrinfo()` it is always threadsafe.
vszakats
force-pushed
the
getaddrinfo-confusion
branch
from
ebded4c to
f21aa1a
Compare
vszakats
removed
hacktoberfest-accepted
feature-window
Curl_getaddrinfo_ex()gets defined withHAVE_GETADDRINFOset. But,hostip4.cused it withHAVE_GETADDRINFO_THREADSAFEset alone. It meant a build with the latter, but without the former flag could result in calling this function but not defining it, and failing to link.Patch this by adding an extra check for
HAVE_GETATTRINFOaround the call.Before this patch, build systems prevented this condition. Now they don't need to, so delete the build-system-specific logic from CMake.
Follow-up to 67d8862
Closes #xxxx