Opened 7 years ago
Closed 7 years ago
#9496 closed enhancement (fixed)
cmake-3.9.0
| Reported by: | Owned by: | ||
|---|---|---|---|
| Priority: | normal | Milestone: | 8.1 |
| Component: | BOOK | Version: | SVN |
| Severity: | normal | Keywords: | |
| Cc: |
Description
New minor version.
Change History (3)
comment:1 by , 7 years ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
comment:2 by , 7 years ago
Working fine, but the testsuite has been doing my head in. Fortunately, I ignore testsuites for almost everything in BLFS, unless I'm editing (it wouldn't be the first time if all existing tests worked, but a change in the current version made it unuseful).
On first run of the tests, yesterday, I got one failure - LinkStatic. Well duh, I hide almost all of my static libs, so that wasn't unexpected (I didn't know it would test that).
So, I made sure that libm.a (which it uses) and libc.a were available (libc_nonshared.a is always available). No change. So I knocked up a script to make ALL static libs (except those few that I've deleted) available - no joy.
Then I looked at the test - a comment suggests it fails if the shared libm is used, so I tried hiding that - bad move, many failures and that test was amongst them.
Today, I've been trying to manually link that test, and totally failing (why are the commands to link test progs often so hard to find ?). So I gave it another attempt (fresh build, all static libs available).
I'd hoped that by just providing the test name, it might only run that test. But it ran them all - this time, 163 LinkStatic passed, but 218 - CTestTestStopTime failed (it timed out on a system that was otherwise idle apart from a browser). That one worked fine a couple of times yesterday, so just one of those things.
After that test run had started, and was clearly running through all the tests, I gurgled a bit more and found that bin/ctest -V might provide more info. Trying that now on 3.8.2, it provides a lot of the usual cmake noise about scanning dependencies, building, etc but it does appear to show the commands with a bit more detail. The log is, of course, massively bigger. and even with -j4 -O the output from each test (numbered) is at times mixed up. But those numbers do not seem to match the test numbers, so in the absence of any failures (and more particularly any error messages for linkage failures) I cannot recommend that.
Totally unrelated to the testing: if MAKEFLAGS have been exported (e.g. as -j4 -O) even the bootstrap will run using 4 cores.
Release Notes at https://cmake.org/cmake/help/v3.9/release/3.9.html
I'll give this a try in my forthcoming build.