CFAST compiled with gnu_win can't finish any input file computation

[bewygs]

Hi,

I found a bug when compiling CFAST from source code on windows with gfortran where it can't launch any input file. I use the DOE201 verification example as input file but it won't work for any input file.

Steps to reproduce the error (Windows cmd):

git clone https://github.com/firemodels/cfast.git
cd cfast\Build\CFAST\gnu_win
.\make_cfast.bat
copy ..\..\..\Verification\DOE_Guidance_Report\DOE201.in . # copying any input file will reproduce the error
cfast7_win.exe DOE201.in

The computation never finishes (I have a CI run in my project where it hits the github actions time limit). With the Intel compiler, I do not encounter this bug.

My spec :

• OS: Win 11
• CFAST version: latest commit
• gfortran version: 13.2.0 (MinGW-W64 x86_64-ucrt-posix-seh, built by Brecht Sanders, r8)

Workaround:
If I edit the gnu_win section in the Makefile and compile with the optimization flag -O0 instead of -O2, the issue disappears and DOE201 runs normally. I can open a PR to adress that.

[gforney]

thanks for the report. I'll take a look

[mcgratta]

Glenn -- can you run the case first on linux using the Gnu compiler?

[gforney]

I'm not seeing the problem on spark our linux computer (gfortran 11.5). I am seeing the problem on windows (gfortran 13.2) , an intel mac (gfortran 15.2) and am arm mac (gfortran 15.2). in all 3 cases changing O2 to O0 causes cases to run

[mcgratta]

With -O2, what happens when you run CFAST from the command prompt?

[gforney]

on windows it hangs doesn't give any output on intel mac get this ***@***.*** Visualization (glutfix) $ ../../../cfast/Build/CFAST/gnu_osx/cfast7_osx cfast_test.in ***Error, 100Consecutive time steps with size below 0.1000E-08 at t = 0.9644E-02 on arm mac on case hangs like on windows a second case gives a DASSL error (indicating cfast rhs routine it is calling is not giving good information) ***@***.*** Visualization (glutfix) $ ../../../cfast/Build/CFAST/gnu_osx/cfast7_osx cfast_target.in ***Error, at t (=r1) and stepsize h (=r2) the corrector failed to converge repeatedly or with abs(h)=hmin ierror,r1,r2 = 650 0.1449D+00 0.9490D-08 Solution component with the greatest error is pressure in room 1 ***Error, dassl - idid = -7

…

On Mon, Feb 23, 2026 at 11:54 AM Kevin McGrattan ***@***.***> wrote: *mcgratta* left a comment (firemodels/cfast#2231) <#2231 (comment)> With -O2, what happens when you run CFAST from the command prompt? — Reply to this email directly, view it on GitHub <#2231 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AC6UCRTEI3HRKUCPJXSPZEL4NMWCRAVCNFSM6AAAAACV4P5FHOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTSNBVHE4DONBRGA> . You are receiving this because you were assigned.Message ID: ***@***.***>

-- Glenn Forney

[mcgratta]

For windows, can you pepper a few print statements throughout the main routine? In FDS, I use VERBOSE=T print statement in main.f90 to see if the simulation even gets off the ground and where it hangs.

[rpeacoc]

There is an equivalent in CFAST. I’ll have to look to remember what the command input isSent from my iPadOn Feb 26, 2026, at 9:10 AM, Kevin McGrattan ***@***.***> wrote:mcgratta left a comment (firemodels/cfast#2231) For windows, can you pepper a few print statements throughout the main routine? In FDS, I use VERBOSE=T print statement in main.f90 to see if the simulation even gets off the ground and where it hangs. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[preneke]

Add &DIAG DASSEL_DEBUG_PRINT = ‘ON’ DEBUG_PRINT = ‘on’ RESIDUAL_DEBUG_PRINT =’ON’/ This will give you new files. One will ‘inputfilename_resid.csv’ could be large but it will tell you if the solver is actually getting a first solution. The second will be named ‘inputfilename_diagnostics.csv’ and the last will give you ‘inputfilename.debug.’ The should give you a lot of information on where things are failing, along with the ‘inputfilename.log’ From: Richard Peacock ***@***.***> Sent: Thursday, February 26, 2026 9:27 AM To: firemodels/cfast ***@***.***> Cc: Subscribed ***@***.***> Subject: [EXTERNAL] Re: [firemodels/cfast] CFAST compiled with gnu_win can't finish any input file computation (Issue #2231) [https://avatars.githubusercontent.com/u/12814856?s=20&v=4]rpeacoc left a comment (firemodels/cfast#2231)<#2231 (comment)> There is an equivalent in CFAST. I’ll have to look to remember what the command input isSent from my iPadOn Feb 26, 2026, at 9:10 AM, Kevin McGrattan ***@***.***<mailto:***@***.***>> wrote:mcgratta left a comment (firemodels/cfast#2231) For windows, can you pepper a few print statements throughout the main routine? In FDS, I use VERBOSE=T print statement in main.f90 to see if the simulation even gets off the ground and where it hangs. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***<mailto:***@***.***>> — Reply to this email directly, view it on GitHub<#2231 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADDV4MDYBWXW6FNHMPYUPCT4N37CXAVCNFSM6AAAAACV4P5FHOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTSNRWHE3TSOJRGE>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>>