-
-
Notifications
You must be signed in to change notification settings - Fork 3.6k
Build suddenly failing (new GCC issue) #3185
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
It looks like gcc is being killed:
I'm not sure under what causes this -- it could be part of our build limits like the memory limit, but that usually appears in a more obvious way. Can you branch off of a working commit (last commit is 9acb311a) and see if the build works? If it doesn't, it might be a matter of our last image deploy -- though timeframe doesn't quite line up there. |
Sure. Is there a way to speed up this process? In my experience, it takes
RTD up to 24 hours to see that a new branch has been added to a project.
On Wed, Oct 25, 2017 at 10:28 AM Anthony ***@***.***> wrote:
It looks like gcc is being killed:
gcc: internal compiler error: Killed (program cc1plus)
I'm not sure under what causes this -- it could be part of our build
limits like the memory limit, but that usually appears in a more obvious
way.
Can you branch off of a working commit (last commit is 9acb311a) and see
if the build works? If it doesn't, it might be a matter of our last image
deploy -- though timeframe doesn't quite line up there.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#3185 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGHnH9sCY2REdxGFUYIcDThGIvkM_FR9ks5sv2yMgaJpZM4QFJaN>
.
--
Kevin Thornton
Associate Professor
Ecology and Evolutionary Biology
UC Irvine
http://www.molpopgen.org
http://github.com/ThorntonLab
http://github.com/molpopgen
|
RTD should pick up on the new branch if you have a webhook set up. You can check this in your project admin dashboard, under integrations. If it does have a webhook setup, there might be something else happening |
I do have the webhook set up. I just deleted and set up a new one. My point is that the webhooks work, but the process is very slow. I often have to come back the next day if I'm trying to test building a new branch. This slowness has been going on for a really long time. I figured this was expected. |
Here is the output from the webhook from pushing a new copy of a branch based on the last commit to work on RTD. There is no indication of error, but the new branches do not show up under "Versions", but I expect they will in several hours, or possibly as late as tomorrow. The link is https://readthedocs.org/dashboard/fwdpy11/integrations/16694/exchange/8527d289-d1d1-4cf1-a4a9-201244aae761/ The output is:
And
|
Nope, expected behavior should be that the branches appear once we receive the webhook. A delay either indicates a malfunctioning/missing webhook, or a bug on our version creation code. I seem to recall another report of this which was tied to the cloned repository on our build server. I'll open a separate issue for this. |
This should perhaps be a separate issue, but the docs for webhooks could use some updating. For example this text appears to be incorrect:
No such directions exist.
|
Regarding the original bug report, does building the previous commit work as expected? |
I cannot check that until the webhook issue is fixed. It has been about 72
hours and RTD does not see the new branch.
On Fri, Oct 27, 2017 at 5:58 PM Anthony ***@***.***> wrote:
Regarding the original bug report, does building the previous commit work
as expected?
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#3185 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGHnHypad5RYBqjTavdadcqxFWFHYr6hks5swnuHgaJpZM4QFJaN>
.
--
Kevin Thornton
Associate Professor
Ecology and Evolutionary Biology
UC Irvine
http://www.molpopgen.org
http://github.com/ThorntonLab
http://github.com/molpopgen
|
One thing that would be helpful is if RTD provided an example of a Python extension requiring compilation and linking to other C/C++ libraries installed via Conda. I have brought this up before in other issues, but that request has never been addressed. Many of the existing docs are quite vague in how to set up such projects for RTD integration, and I only got things running via help from another user who had figured it out. For examples, #2707 and #2958 were never addressed, and the latter is a somewhat serious problem, IMO. Further, I know of other packages where the maintainer gave up on trying to build on RTD and instead mocked all the docs, due to lack of documentation in how to set things up. The project of mine is pretty thoroughly tested on Travis. All the builds that are failing on RTD pass on Travis, using both Anaconda GCC and several combos of Ubuntu GCC 5 and 6 with various Python versions (https://travis-ci.org/molpopgen/fwdpy11/builds/291754232). |
OK, I was able to get a branch built based on the last commit successfully built on RTD. It did indeed pass: https://readthedocs.org/projects/fwdpy11/builds/6244258/ I guess I should look at what the next commit did... |
What GCC version is RTD using? I notice that the failing builds are compiling with C++14, implying that a GCC >= 5.something is being used. Further, when I attempt to install GCC from conda (which is 4.8.5), builds fail with no meaningful error. For this build, the only change was to add GCC as dependency in the environment file for RTD: |
For reasons I do not understand, build using conda GCC suddenly started to pass within the last hour. This commit was all it took: molpopgen/fwdpy11@39ab23b Results are here: https://readthedocs.org/projects/fwdpy11/builds/6246693/ The change uses conda-forge gcc, which is 4.8.5. The implication is that there is a config issue with the RTD GCC that is interacting badly with recent versions of my project. Importantly, this interaction cannot be attributed to a problem with the code in my project--all builds pass on conda GCC, GCC5, and GCC6 on Travis/Linux as well as several other systems where we use and test the code. Edit: For example, here's the recent build results on Travis: https://travis-ci.org/molpopgen/fwdpy11. That's presumably spanning whatever RTD is using? |
RTD uses this docker images to build your docs https://github.com/rtfd/readthedocs-docker-images/tree/releases/3.x. Anyway, looks like your problem is solved, right? |
Closing this as the user hasn't responded or given more information, feel free to reopen with more information if you are still having this problem :). |
Details
Expected Result
The Python extension module should build, and then the manual should build.
Actual Result
GCC fails with an internal compiler error. This is new behavior. Also, attempting to install GCC via conda results in failure to link errors.
The text was updated successfully, but these errors were encountered: