Perfectly clear now. Thank you for the hint Bruno :)! [email protected] schrieb am Freitag, 23. Januar 2026 um 16:40:00 UTC+1:
> Dear Magdalena, > If you want to have unit tests that use restart files, the restart files > need to be generated with the same version of p4est that is used to run the > restart. As far as I know, p4est checkpoints are not compatible (forward > nor backward compatibility does not seem to be there). > > So for example, in Lethe, we have unit tests that actually restart a quite > massive simulation, but only to carry out one time-step. Generating the > restart file is time consuming (say 30 minutes), but the test itself runs > in less than a minute. However, we store the checkpoint information as an > artifact on our repository. So if the CI of Lethe, which uses the deal.II > docker image, does not have the same version of p4est as the person > generating the test, then these test will always crash. > > The best workaround would be to generate the restart files "in-situ". > Having a first step in the CI process that generates the restart files. The > issue with this is that generating the restart files must then be > relatively fast, which is not always the case for us right now. > Asymptotically, we will most likely have an entirely different CI step for > restarts. > > I hope that explains the potential issue :). > Cheers > BB > > > > On Friday, January 23, 2026 at 2:40:57 a.m. UTC-5 [email protected] > wrote: > >> Dear Bruno, >> >> thank you for the detailed explanation. Although it seems to be still a >> mess, at least it will help us to work around possible issues :-). >> >> What do you mean by `your restart files won't work`? Because there is an >> incompatibility between the p4est versions or what is the issue there? >> >> Best, >> Magdalena >> >> >> >> [email protected] schrieb am Freitag, 23. Januar 2026 um 03:23:39 UTC+1: >> >>> Dear Magdalena, >>> >>> Bruno does not really know the rationale, regrettably. I wish I was >>> smarter, but sometimes these things still confuse me :). >>> If you use a very modern GCC version, which Manjaro uses (I have the >>> same issue with Ubuntu 25.10), then the compilation breaks when it tries to >>> use BLAS. So you should not specify that BLAS is not needed. However, under >>> older compiler version (say GCC with Ubuntu 24.03), then the reverse occurs >>> and if you don't specify the "without blas" then it won't compile. So yeah, >>> this is a major headache. >>> >>> I have proposed a new PR online which fixes this using a modern version >>> of p4est. https://github.com/dealii/candi/pull/436 >>> There is a caveat though. The deal.II docker image uses p4est 2.3.6, so >>> if you are using an online continuous integration with that image, your >>> restart files won't work. >>> Nothing is easy with these things it seems :). >>> >>> Cheers >>> BB >>> >>> On Thursday, January 22, 2026 at 4:44:34 a.m. UTC-5 >>> [email protected] wrote: >>> >>>> Dear Gerald, >>>> >>>> I am encountering the same issue with p4est 2.3.6 (the version used in >>>> candi). I am building on Manjaro with GCC 15.2.1 (20250813). It could be >>>> that this problem did not occur with earlier compiler versions (as I >>>> definitely compiled p4est successfully couple of months ago). >>>> >>>> The issue appears to be related to the p4est configuration used by >>>> deal.II, specifically here: >>>> >>>> https://github.com/dealii/dealii/blob/3ab62bd61ec7ac3221bcca757474cf2bc3c520a8/doc/external-libs/p4est-setup.sh#L120 >>>> >>>> In this script, deal.II explicitly forces p4est to be built without >>>> BLAS support. I do not fully understand why this is required (perhaps >>>> Bruno >>>> knows the rationale), but if I remove this option, the build succeeds. >>>> This >>>> makes me suspect a faulty or outdated implementation in p4est related to >>>> the --without-blas flag, which may be exposed by newer compilers. >>>> >>>> Best, >>>> Magdalena >>>> >>>> [email protected] schrieb am Mittwoch, 7. Januar 2026 um 15:15:38 >>>> UTC+1: >>>> >>>>> Dear BB, >>>>> >>>>> Thank you for the email. >>>>> >>>>> I'm still on my way back home from a holiday trip. I'll give your >>>>> advice a try as soon as I have access to me machine. >>>>> >>>>> Bye for now, >>>>> >>>>> Gerald >>>>> >>>>> Gerald Marewo >>>>> Cell: +27(71)705 0830 <+27%2071%20705%200830>, Email: >>>>> [email protected] >>>>> >>>>> >>>>> >>>>> On Mon, 5 Jan 2026, 10:35 [email protected], <[email protected]> >>>>> wrote: >>>>> >>>>>> Dear Gerald, >>>>>> A quick question, did you clean your candi installation folder after >>>>>> installing BLAS and restarting the process? Just to be sure. Sometimes >>>>>> if >>>>>> you do not clean the install and add new libraries, the new libraries >>>>>> are >>>>>> not detected adequately. >>>>>> Otherwise, I am not sure what is the exact issue right now. I have >>>>>> never encountered this problem. >>>>>> If you try to compile p4est as a standalone library on the side, do >>>>>> you get the same error? >>>>>> >>>>>> >>>>>> On Saturday, December 27, 2025 at 10:23:08 a.m. UTC+1 >>>>>> [email protected] wrote: >>>>>> >>>>>>> Dear BB and others, >>>>>>> >>>>>>> I now know that I have BLAS installed: >>>>>>> >>>>>>> (base) gerald@hephaestus:~$ dnf search blas-devel >>>>>>> Updating and loading repositories: >>>>>>> Repositories loaded. >>>>>>> Matched fields: name (exact) >>>>>>> blas-devel.i686 BLAS development libraries >>>>>>> *blas-devel*.x86_64 BLAS development libraries >>>>>>> Matched fields: name >>>>>>> flexiblas-devel.i686 Development headers and libraries for >>>>>>> FlexiBLAS >>>>>>> flexiblas-devel.x86_64 Development headers and libraries for >>>>>>> FlexiBLAS >>>>>>> hipblas-devel.x86_64 Libraries and headers for hipblas >>>>>>> liblas-devel.i686 libLAS development files >>>>>>> liblas-devel.x86_64 libLAS development files >>>>>>> openblas-devel.i686 Development headers and libraries for OpenBLAS >>>>>>> *openblas-devel*.x86_64 Development headers and libraries for >>>>>>> OpenBLAS >>>>>>> rocblas-devel.x86_64 Libraries and headers for rocblas >>>>>>> xblas-devel.i686 Development files for xblas >>>>>>> xblas-devel.x86_64 Development files for xblas >>>>>>> >>>>>>> Attempting to install deal-candi fails at second attempt with the >>>>>>> same error as reported before. >>>>>>> >>>>>>> Bye for now. >>>>>>> >>>>>>> Gerald. >>>>>>> On Friday, 26 December 2025 at 12:00:05 UTC+2 Gerald Marewo wrote: >>>>>>> >>>>>>>> Dear BB, >>>>>>>> >>>>>>>> Thank you for your reply. Let me take a shot at installing BLAS. >>>>>>>> >>>>>>>> Bye for now >>>>>>>> >>>>>>>> Gerald Marewo >>>>>>>> Cell: +27(71)705 0830 <+27%2071%20705%200830>, Email: >>>>>>>> [email protected] >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> On Fri, 26 Dec 2025, 11:54 [email protected], <[email protected]> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> Dear Gerald, >>>>>>>>> It does not seem like an out-of-memory error since the compilation >>>>>>>>> of p4est is clearly not a memory intensive operation (it's only C >>>>>>>>> after >>>>>>>>> all). >>>>>>>>> Looking at your error log, this seems to be where compilation >>>>>>>>> crashed: >>>>>>>>> In file included from >>>>>>>>> /home/gerald/bin/dealii-candi/tmp/unpack/p4est-2.3.6/sc/src/sc_dmatrix.h:32, >>>>>>>>> from >>>>>>>>> /home/gerald/bin/dealii-candi/tmp/unpack/p4est-2.3.6/sc/src/sc_dmatrix.c:24: >>>>>>>>> /home/gerald/bin/dealii-candi/tmp/unpack/p4est-2.3.6/sc/src/sc_dmatrix.c: >>>>>>>>> >>>>>>>>> In function ‘sc_dmatrix_add’: >>>>>>>>> /home/gerald/bin/dealii-candi/tmp/unpack/p4est-2.3.6/sc/src/sc_blas.h:137:33: >>>>>>>>> >>>>>>>>> error: too many arguments to function ‘sc_blas_nonimplemented’; >>>>>>>>> expected 0, >>>>>>>>> have 6 >>>>>>>>> 137 | #define SC_BLAS_DAXPY (void) sc_blas_nonimplemented >>>>>>>>> | ^~~~~~~~~~~~~~~~~~~~~~ >>>>>>>>> /home/gerald/bin/dealii-candi/tmp/unpack/p4est-2.3.6/sc/src/sc_dmatrix.c:627:5: >>>>>>>>> >>>>>>>>> note: in expansion of macro ‘SC_BLAS_DAXPY’ >>>>>>>>> 627 | SC_BLAS_DAXPY (&totalsize, &alpha, X->e[0], &inc, >>>>>>>>> Y->e[0], &inc); >>>>>>>>> | ^~~~~~~~~~~~~ >>>>>>>>> >>>>>>>>> >>>>>>>>> What I would look into are the following: >>>>>>>>> - Did you install all of the recommended dependencies from Candi >>>>>>>>> before proceeding with the installation? It seems as if you do not >>>>>>>>> have an >>>>>>>>> installation of BLAS on your computer, which would prevent p4est from >>>>>>>>> actually compiling. >>>>>>>>> - If BLAS is indeed installed correctly, then I would try to >>>>>>>>> install a more recent version of p4est. This could be achieved my >>>>>>>>> changing >>>>>>>>> the version and the checksum in >>>>>>>>> the deal.II-toolchain/packages/p4est.package file. >>>>>>>>> >>>>>>>>> However, from what it seems right now is that you are missing a >>>>>>>>> BLAS library on your computer. >>>>>>>>> Don't hesitate if you have more questions >>>>>>>>> Cheers >>>>>>>>> BB >>>>>>>>> >>>>>>>>> >>>>>>>>> On Sunday, December 21, 2025 at 11:20:15 a.m. UTC+1 >>>>>>>>> [email protected] wrote: >>>>>>>>> >>>>>>>>>> Dear all, >>>>>>>>>> >>>>>>>>>> I recently tried to install deal.II using candi on Fedora Linux >>>>>>>>>> 43 (Workstation Edition) and I got the attached error message. >>>>>>>>>> >>>>>>>>>> I followed the instructions given when I ran the candi shell >>>>>>>>>> script as follows: >>>>>>>>>> >>>>>>>>>> ./candi.sh -j8 --prefix=/home/gerald/bin/dealii-candi >>>>>>>>>> >>>>>>>>>> Could it be that my machine ran out memory? I will try again >>>>>>>>>> later without the -j8 option. >>>>>>>>>> >>>>>>>>>> Could you please assist me with seeing what is going wrong >>>>>>>>>> >>>>>>>>>> I hope to hear from you soon. >>>>>>>>>> >>>>>>>>>> Regards, >>>>>>>>>> >>>>>>>>>> Gerald. >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> -- >>>>>>>>> The deal.II project is located at http://www.dealii.org/ >>>>>>>>> For mailing list/forum options, see >>>>>>>>> https://groups.google.com/d/forum/dealii?hl=en >>>>>>>>> --- >>>>>>>>> You received this message because you are subscribed to a topic in >>>>>>>>> the Google Groups "deal.II User Group" group. >>>>>>>>> To unsubscribe from this topic, visit >>>>>>>>> https://groups.google.com/d/topic/dealii/V_MjMXOEjX8/unsubscribe. >>>>>>>>> To unsubscribe from this group and all its topics, send an email >>>>>>>>> to [email protected]. >>>>>>>>> To view this discussion visit >>>>>>>>> https://groups.google.com/d/msgid/dealii/b99db8d5-4997-4c20-9855-03071a381871n%40googlegroups.com >>>>>>>>> >>>>>>>>> <https://groups.google.com/d/msgid/dealii/b99db8d5-4997-4c20-9855-03071a381871n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>>>>>> . >>>>>>>>> >>>>>>>> -- >>>>>> The deal.II project is located at http://www.dealii.org/ >>>>>> For mailing list/forum options, see >>>>>> https://groups.google.com/d/forum/dealii?hl=en >>>>>> --- >>>>>> You received this message because you are subscribed to a topic in >>>>>> the Google Groups "deal.II User Group" group. >>>>>> To unsubscribe from this topic, visit >>>>>> https://groups.google.com/d/topic/dealii/V_MjMXOEjX8/unsubscribe. >>>>>> To unsubscribe from this group and all its topics, send an email to >>>>>> [email protected]. >>>>>> >>>>> To view this discussion visit >>>>>> https://groups.google.com/d/msgid/dealii/261d9ad4-76d9-4754-8b93-84ac96a668b4n%40googlegroups.com >>>>>> >>>>>> <https://groups.google.com/d/msgid/dealii/261d9ad4-76d9-4754-8b93-84ac96a668b4n%40googlegroups.com?utm_medium=email&utm_source=footer> >>>>>> . >>>>>> >>>>> -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/dealii/80a38789-18e5-485b-b456-ed714985232fn%40googlegroups.com.
