-
Notifications
You must be signed in to change notification settings - Fork 30
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
Build failure on Debian unstable #19
Comments
Hm. I've built it with exactly that clang version (6.0.1-9) several times (on debian unstable) and I tried it just now again just to be sure. I just run:
And it works fine for me. I use separate checkouts/builds for all components for the fuzzer. |
On a different machine, also latest Debian unstable and also a dev machine, I tried with a fresh clone of curl-fuzzer in /tmp.
|
Just a guess... my 'make' is an alias to 'make -j$(nproc)'. Is there a problem with parallel builds ? Will check later, have to finish some work now :-| |
I don't believe there's a problem with parallel builds but obviously there could be something subtle going on. Make's usually pretty good at resolving dependencies in order. libstandaloneengine.a should be built for you if required; it's defined in the Automake at https://github.com/curl/curl-fuzzer/blob/master/Makefile.am#L61 and this generally works on the travis runners and such. I'd retry with a simple |
In curl-fuzzer/ a |
For me it looks like a missing dependency. I can see |
This fixes it for me - thanks for for your time. |
BTW, I now see memory leaks in curl - any interest in reporting those ? Or it that expected behavior ? |
@rockdaboot: I'm surprised that that patch works, because Could you make a similar |
NP |
So, from the broken log:
From the working log:
To my eyes those are identical, so I'm at a loss as to why one of them is working and one of them is not... did you do a clean build for the second one? |
Yes I did. I can tell you that the broken build didn't generate libstandaloneengine.a at all. You can see that in I now revert to the previous commit (hard resetting my changes) and try again with Though this looks good:
|
If I replace in Makefile.am
then Reverting the above brings back the issue. Looks like LIB_FUZZING_ENGINE is set outside the of 'make' - but I checked my environment, no such variable. |
As a PoC, I changed |
So the idea of Possibly there's some easy way around this. Perhaps a configuration option could be used instead to set this up. |
If not setting But the way LIB_FUZZING_ENGINE ?= is used in Makefile.am looks absolutely fine. Also, renaming that to show there is no environment issue and then using just = with success tells me that there must be a make issue (make 4.2.1-1.2). I did those changes directly in the generated Makefile to eliminate possible automake issues. BTW, |
|
Not sure what is going on, clang is 6.0.1-9 and used often for fuzzing other projects (also for OSS-Fuzz).
I naively followed the instructions (../curl was fetched from upstream a minute ago)
Possibly something is missing, but not to easily detectable from the above messages.
I appreciate any help / ideas :-)
The text was updated successfully, but these errors were encountered: