PyREC scaffolding (from YREC development, 7be4133)/ parmin.fGlossary

parmin.f

1541 lines · 2 definitions · on GitHub (7be4133)

Python version: parmin.py, 961 lines, on the right (switch to Python above the code).

Variables on this page (362)

abstol, acfpft, alatm_alpha, alatm_feh, alphac, alphae, alpham, atime, atmbeg, atmd0, atmerr, atmmax, atmmin, atmstp, betac, c_2, calsolage, calsolzx, cc13, cc23, cln, clsun, cmin, cmixla, codm, compacc, constfactor, creim, crsun, cstdiffmix, cstmixing, djok, dmdt0, dpenv, dt_gs, dtdif, dtwind, end_dcen, end_xcen, end_ycen, endage, envbeg, enverr, envmax, envmin, envstp, etadh0, etadh1, excen, exl, exm, exmd, expr, exr, extau, exw, fc, fcore, fcorr0, fcorri, fczdmdt, fes, fesc, fgry, fgrz, fgsf, fgsfc, flag_dw, flag_dx, flag_dz, fmu, fo, frac_c, frac_n, frac_o, fss, fssc, ftotdmdt, fw, grtol, hpttol, htoler, ialex06, iclcd, icondopacp, idebug, ies, ifermi, ifirst, igsf, ilambda, ilast, iliv95, imax, imilne, imodpt, impjmod, imu, inewcp, ioatm, ioatma, iopale, iopatm, iopenv, iopmod, iopurez, iowr, ipver, iresca, irun, iscvh, iscvhe, iscvz, isetiso, isetmix, ishort, isnu, isound, istand, istor, itdif1, itdif2, itfp1, itfp2, itrack, itrver, kemmax, kttau, kttau0, l_zerodt_prems, lacout, ladov, lalex06, laltptau100, lbscale, lcalcenv, lcals, lcalsolzx, lcalst, lcodm, lcondopacp, lcore, lcorr, ldebug, ldh, ldifad, ldifli, ldify, ldifz, ldisk, lendag, lenvg, lexcom, lfirst, linstb, lisotope, ljdot0, lkuthe, lmdot, lmilne, lmixture, lmwind, lnew0, lnewcp, lnewdif, lnewnuc, lnews, lnewtcz, lnewvars, lnoj, lnulos1, lnumderiv, lopal95, lopale06, lovmax, lovstc, lovste, lovstm, lphhd, lpout, lptime, lpulse, lpurez, lreimer, lrel, lrossby, lrot, lrwsh, lscv, lsdepth, lsemic, lsenv0a, lsetdt, lsnu, lsolid, lsolwind, lstatm, lstch, lstenv, lstmod, lstore, lstpch, lstphys, lstrot, lstruct_time, lteff, lthoul, lthoulfit, ltrack, ltrist, lttau, lvfc, lwnew, lxbe91, lxbe92, lxbe93, lxli6, lxli7, lzramp, mcore, niter1, niter2, niter3, niter4, niter_gs, nmodls, npenv, npoint, nprtmod, nprtpt, numrun, nuse, optol, pdisk, pmma, pmmb, pmmc, pmmd, pmmjd, pmmm, pmmmd, pmmsolp, pmmsoltau, pmmsolw, poa, pob, poc, pomax, r12_13, r14_15, r16_17, r16_18, rcrit, reltol, rescal, rsclzc, rsclzm1, rsclzm2, s0_1_1, s0_1_12, s0_1_13, s0_1_14, s0_1_15_c12alp, s0_1_15_o16, s0_1_16, s0_1_be7e, s0_1_be7p, s0_3_3, s0_3_4, s0_hep, s0_pep, s0p_1_1, s0p_1_12, s0p_1_13, s0p_1_14, s0p_1_16, s0p_1_be7p, s0p_3_3, s0p_3_4, s0pp_1_12, s0pp_1_13, s0pp_1_16, s0pp_1_be7p, sbe91, sbe92, sbe93, senv0a, setdt, sli6, sli7, spotf, spotx, sr, sstandard, steff, stolr0, tcut, tdisk, tenv0, tenv1, tgcut, time_core_min, time_dl, time_dp, time_dr, time_dt, time_dw_global, time_dw_mix, time_dx_core_frac, time_dx_core_tot, time_dx_shell, time_dx_total, time_dy_core_frac, time_dy_core_tot, time_dy_shell, time_dy_total, time_max_dt_frac, tmolmax, tmolmin, tol_czbase_fine_width, tol_dl_max, tol_dm_max, tol_dm_min, tol_dp_core_max, tol_dp_czbase_max, tol_dp_env_max, tol_dx_max, tol_dz_max, toll, tollaol, tolr, tolz, tridl, tridt, tscut, vnew, walpcz, weakscreening, wmax, wnew, xb10_ini, xb11_ini, xbe9_ini, xenv0a, xh2_ini, xhe3_ini, xli6_ini, xli7_ini, xls, xlstol, xmin, xnewcp, xsbe91, xsbe92, xsbe93, xsli6, xsli7, ymin, zenv, zenv0a, zopal951, zxmix

Purpose

PARMIN reads a run's parameters and sets up its input and output. It holds the defaults of a few user parameters (most defaults are now initializers in the modules), reads the two namelist files (CONTROL, usually yrec8.nml1, and PHYSICS, usually yrec8.nml2), expands placeholders in file names, derives some parameters from others, opens most output files and several tables, writes the run's settings to the .short file, and interprets the "run cards" (the sequence of evolve and rescale steps). It is the code's single place of configuration.

How it fits

  • Inputs: the two namelist files, named on the command line (GETARG(1) and GETARG(2); JVS 02/11) or defaulting to yrec8.nml1 and yrec8.nml2; the code's version and git hash, YRECVER and GITHASH (arguments, from SETVERSION through the main program; formerly /VERSION/); the environment variables YREC_INPUT, YREC_START and YREC_OUTPUT, if set. GETARG, getenv and system (line 758↗, line 1527↗, line 828↗) are compiler extensions; standard Fortran has GET_COMMAND_ARGUMENT, GET_ENVIRONMENT_VARIABLE and EXECUTE_COMMAND_LINE.
  • Who calls it: the main program in main.f, once, first.
  • What it calls: EXPAND_VALUE (in this file), the shell through system (mkdir -p for the output folder), and REMAP, which maps the newer, "more intuitively named" parameters (MHP 6/14, 10/14: TIME_…, TOL_…, FLAG_…) onto the older vectors ATIME, HPTTOL and the like.
  • Outputs:
  • Arguments: 12 table file names (FALEX06, FALLARD, FATM, FFERMI, FLIV95, FPATM, FPENV, FPMOD, FPUREZ, FSCVH, FSCVHE, FSCVZ), which the main program passes on to SETUPS and other set-up routines. Since MHP 8/25, file names travel as arguments rather than in COMMON blocks, "as character strings in common blocks can cause problems."
  • Modules: 11, almost all of them for the run's parameters. The namelists read straight into module variables: NMLPAR_MOD holds those the physics routines read, NMLBYPASS_MOD those that only set others (the cross sections, the TIME_… and TOL_… parameters for REMAP, the wind-law indices), and CONSTANTS_MOD (CLSUN, CRSUN, and CLN, CC13, CC23, C_2 set here) and INTPAR_MOD (STOLR0, IMAX, NUSE) a few more. The others: IOUNITS_MOD (every unit number), RUNLOOP_MOD (the run cards: RESCAL, IRESCA, LENDAG, LSETDT), COMP_MOD (ZENV, bug 3), CCOUT_MOD (LPHHD), ACDPTH_MOD (LACOUT), LABEL_MOD and NEWOPAC_MOD. The site's modules.html lists every module and its users.
  • Files opened: the .short diagnostic file (ISHORT, unit 20), the stored models (ISTOR, with a header if LSTORE), the first model (IFIRST, read), the last model (ILAST), the per-shell file (IMODPT), and as switched on: the track, solar neutrinos, debugging, Milne variables (appended), pulsation files, the conductive opacities, the OPAL 2006 EOS table, the acoustic-depth file (…calcad, named after the .short file) and, always, the sound-speed file FSOUND (default Csound.dat, unit ISOUND), which WRTMOD writes when LSOUND is on. Before that it creates the output folder, the path of FTRACK up to its last /.
  • Unit numbers: fixed here, from IOWR = 6 (the screen or status file) to 90 (IALEX06); for example the OPAL EOS table is 49, and ISOUND is 500.

What it does, in order.

  1. Defaults. Most are now initializers in the modules (NMLPAR_MOD, NMLBYPASS_MOD, CONSTANTS_MOD, INTPAR_MOD); PARMIN's own DATA keep those of the run cards, the mixture table, ALFA, FK, WMAX_SUN, AWIND, FSOUND and blank file names. Among the defaults, as their comments state: the cross sections of Solar Fusion II (Adelberger et al. 2011) and their derivatives (S0_…, S0P_…, S0PP_…); Li and Be destruction cross sections (Fowler et al. 1967, Rolfs & Kavanagh 1986, Fowler 1988); the GS98 CNO fractions and $Z/X = 0.02292$ and the Lodders et al. (2021) isotope ratios and light-element abundances; alternative mixtures AAG21, M22P, M22M (Asplund et al. 2021; Magg et al., given as 2021 in the comment); a G&N93 mixture VNEW for the envelope; $L_\odot = 3.8515\times10^{33}$ erg s⁻¹ and $R_\odot = 6.9598\times10^{10}$ cm; the burning cutoffs TCUT. The mixing length has no default any more: the old CMIXL $=1.4$ never reached the run, which takes CMIXL = CMIXLA(NK) in the main program for each run card.

    The mixtures are Magg et al. (2022). The comment cites 'M22M' and 'M22P' as "MAGG ET AL. 2021 (MET,PHOT) A&A 661,140" (line 515↗), but input/opacity/atomic/README cites "Magg et al. (2022)," as the names suggest. The comment's order (MET,PHOT) is also reversed against the DATA, which lists 'M22P' ($Z/X = 0.0225$) third and 'M22M' ($0.0226$) fourth (lines 516–517↗), and its list of supported mixtures, 'AAG21',M22M','M22P', lacks the quote before M22M (line 506↗).

  2. Namelists. Both files are read and REMAP is called. Then EXPAND_VALUE expands a leading placeholder in each file name (bug 5), and the output folder is created (bug 4).

  3. Derived settings.

  4. Li/Be cross sections given by the user (LXLI6 and so on) become scale factors relative to the defaults, e.g. SLI6 = XSLI6/5.5D3.

  5. Semiconvection together with any overshoot stops the run. Rotation (LROT) with the advection-diffusion transport of angular momentum (LDIFAD) stops it too: as the code's comment says, DADCOEFT rebuilds its coefficients from equal-grid arrays that no routine sets, so the option is not implemented.

  6. Rotation (LROT): WALPCZ is limited to $[-2, 0]$; LSOLID forces IMPJMOD = 1. For the old wind law (LMWIND false), from the index $\alpha$ (ALFA), as the comments derive: $$\dot J = f_K\, 2.036\times10^{33}\,(1.452\times10^{9})^{\alpha}\left(\frac{\dot M}{10^{-14}}\right)^{1-2\alpha/3}\Omega^{1+4\alpha/3}\left(\frac{R}{R_\odot}\right)^{2-\alpha}\left(\frac{M}{M_\odot}\right)^{-\alpha/3}.$$ This is the Kawaler (1988) law (as the documentation names it); $\dot M$ is "in solar masses per year" (fixed at $2\times10^{-14}$ in COWIND and WIND) and $M$, $R$ relative to the Sun's, as their comments say.

    For the Matt et al. law (LMWIND), the indices $a, b, c, d, m$ (PMMA…PMMM), preset by AWIND ('K97': Krishnamurthi 1997, $m = \alpha/3$; 'V13': van Saders & Pinsonneault 2013, $m = 0.22$; 'CUS': from the namelist), give the exponents of $\dot J$, each quantity relative to the Sun: $$a_\Omega = 1 + a - 2am + 4mb,\quad a_R = 2 + 5m - 4mc,\quad a_M = -m,\quad a_L = 1 - 2m,\quad a_P = 2md,\quad e_\mathrm{cen} = md,$$ with $a_\tau = 0$ (the $\tau_\mathrm{cz}$ term is now applied in MWIND), $c_2 = 0.0506$, CONSTFACTOR $= f_K\,J_\odot/\Omega_\odot^{a_\Omega}$, and, if WMAX_SUN $< 1000$, the saturation threshold $\Omega_\mathrm{sat} = $ WMAX_SUN $\times\,\Omega_\odot$.

  7. Composition changes: a single species to rescale (LNEWCP, ANEWCP), a CNO mixture (ISETMIX, AMIX, with checks on a custom one) and isotope ratios (ISETISO, AISO), all applied later to the starting model by STARIN (the documentation's value 2 for a custom mixture does nothing: bug 2).

  8. Settling switches: LDIFZ turns on LDIFY and LTHOUL; LDIFLI turns on LNEWDIF, LDIFY, LDIFZ, LTHOUL, turns off LTHOULFIT, and sets ILAMBDA = 4.
  9. Summary. A line of tolerances and rotation settings is written first, right after the semiconvection and LDIFAD checks. After the derived settings, the PHYSICS and CONTROL namelists are written to ISHORT as read, and the output-only namelist DERIVED with the values derived above (the wind-law exponents, the settling switches, the mixture, LREL, …). Then the mixing lengths are checked: each CMIXLA entry the run uses (only CMIXLA(1) under solar calibration, LCALS, otherwise the first NUMRUN, at most 50) must be positive, or the run stops; so must the envelope triangle's size, TRIDT and TRIDL. Then the atmosphere choice (KTTAU 0–5), and the run's description with the version line (# YREC v…, also written at the head of the .store file and the track).
  10. Run cards (NUMRUN cards, at most 50): KINDRN = 1 evolves NMODLS models (to an age, a central D, X or Y, or with a fixed time step); 2 rescales the model (mass, $X$, $Z$, core mass) and relaxes it; 3 rescales and evolves. LENDAG and LSETDT are set only for evolve cards (lines 1396–1406↗), so the stop conditions (ENDAGE, END_XCEN, END_YCEN, …) are ignored on a rescale-and-evolve run: main.f, suspected bug 1, whose fix belongs here. The message of a rescale-and-evolve card is that of a rescale card (bug 1), and each card sets ZENV (bug 3).

Design

PARMIN is a 1,508-line routine that is mostly comments, declarations and data. A parameter now lives in one place, the module variable the namelist reads into. A few namelist entries are still unused (see the question at step 2).

From YREC's documentation

Control Options:

  • Placeholders. {YREC_INPUT}, {YREC_START}, {YREC_OUTPUT} at the start of a path in the CONTROL namelist are replaced by the environment variable of that name, or by a default: the documentation gives ../../input, ../../input/models and output (in the working directory). PARMIN uses ../../startmodels for {YREC_START} (lines 1515–1516↗), and the repository keeps its starting models in startmodels/ (there is no input/models/), so the documentation's default and the "input/models" of the User Guide and the Quick Start are out of date (settled from the code and the repository, Nov 9, 2026).
  • File names. FFIRST the starting model, FLAST the last converged model, FSTOR the snapshots, FTRACK global quantities against time, FSHORT the numerics of the run; FMODPT, FSNU, FSCOMP are "no longer used" (light elements are now in the track). Two of them are still opened and written: FSNU (opened with LSNU, lines 946–950↗) by the main program (main.f, lines 1209–1229↗), and FMODPT (always opened, line 965↗) by PARMIN's description of the run (line 1364↗), wrtmod.f, line 152↗ and seculr.f, lines 486–498↗ (settled from the code, Nov 9, 2026). Table paths are on the pages of the routines that read them.
  • Read here only. KINDRN, RSCLM, RSCLX, RSCLZ, RSCLCM (the run cards, below); AMIX, AISO, ISETMIX, ISETISO (the mixture, STARIN's page); AWIND, FK, ALFA, WMAX_SUN (the wind law, MWIND's page; the Physics page lists ALFA as "Not used. Legacy exponent for Kawaler (1988) wind law," namelist_physics.md line 311, and the modules page calls ALFA and FK exponents, but ALFA, default 1.5, sets the old law's exponents and CONSTFACTOR when LMWIND is off, and PMMM = ALFA/3 for AWIND = 'K97', the default, lines 1036–1040↗, line 1077↗; FK is a scale factor); LXLI6…LXBE93 with XSLI6…XSBE93, manually set p+⁶Li, p+⁷Li and the three p+⁹Be cross sections (keV b; the documentation labels LXBE92 and LXBE93 as p+⁶Li channels, a slip for ⁹Be).

  • Run cards (Run parameters & Kind cards): each card is a phase, KINDRN(i) = 1 evolve, 2 rescale (zero time step, main sequence or core helium burning), 3 rescale and evolve (pre-main sequence, which has no zero-time-step solution); LFIRST(i) starts from FFIRST rather than the previous phase's model; RSCLM, RSCLCM, RSCLX, RSCLZ, if positive, rescale to that mass, core mass (core helium burning only), $X$, $Z$; CMIXLA(i) is the mixing length to rescale to (or run with); LSENV0A(i) moves the outer fitting point to SENV0A(i), which the Control page calls the log of the fractional envelope mass ($10^{-4}$ standard, $10^{-7}$ thin), but which is $\log_{10}(m_\mathrm{fit}/M_\star)$ of the fitting point, a small negative number (as the User Guide says: $-10^{-4}$ standard, about $2\times10^{-4}$ of the mass outside the fitting point; $-10^{-7}$ thin; STARIN negates a positive value; STARIN's page, step 6); NMODLS(i) stops after that many models (two for a first rescaling is recommended, then an evolve run); ENDAGE, END_DCEN, END_XCEN, END_YCEN stop at an age or a central D, $X$ or $Y$.

From the user documentation (User Guide, Quick Start):

  • Reading the namelists. The two files are read in a fixed order, the first argument as CONTROL (unit ISTAND) and the second as PHYSICS (unit IRUN); the Python wrapper accepts them in either order, the program does not. They are opened by the names given, so a relative name is found from the working directory; the User Guide says both files should be in the executable's directory (userguide.md, line 23).

    The namelists are looked for in the working directory. PARMIN opens the two files by the names given (OPEN(UNIT=ISTAND, FILE=YREC1, …), lines 766–767↗; by default yrec8.nml1 and yrec8.nml2, lines 758–764↗), so a relative name is found from where the run starts, wherever the executable is; the User Guide could say "the working directory."

  • The output folder. PARMIN now creates the folder of FTRACK (mkdir -p) before any output is opened, while the Quick Start still tells the user to mkdir output by hand, or the run fails; the step is no longer needed (bug 4 for the case of a name without /).

Where Namelist Variables Live gives the reasons for the modules' split (on NMLBYPASS_MOD's page) and why CHARACTER variables are passed as arguments: in a COMMON block they force padding against numeric members, which compilers treat differently.

Suspected bugs, checked by running the code

Each bug is marked by what running the code showed: (changes results, large) or (small), the fix moving a run clearly beyond a null change; (stops the run); (wrong output), the fix changing only what is written; (harmless in the runs), within the noise; (never occurred), the code ran but the fault did not; (not reached); (cannot be reached), a guard prevents it; (withdrawn), the suspicion was wrong.

1. (wrong output) Checked by KC (Nov 10, 2026) The wrong message for a rescale-and-evolve card.

(lines 1476–1480↗) A KINDRN = 3 card that starts from the previous run's model prints format 451, "RESCALE THE PREVIOUS RUN'S LAST MODEL.," the message of a rescale-only card; its own format 551 ("RESCALE & EVOLVE THE PREVIOUS RUN'S LAST MODEL.") is commented out. Only the screen and .short message is wrong.

A possible fix:

parmin.f

@@ lines 1474–1483 @@
   550          FORMAT(/1X,'RUN #',I3,
      1         '   RESCALE & EVOLVE THE INPUT MODEL: "FIRST MODEL".')
             ELSE
-               WRITE(IOWR,451) NKIND
-               WRITE(ISHORT,451) NKIND
-C   551          FORMAT(/1X,'RUN #',I3,
-C      1         '   RESCALE & EVOLVE THE PREVIOUS RUN''S LAST MODEL.')
+C 11/09/26 KC via Claude
+C code changed to print the rescale-and-evolve message, format 551.
+               WRITE(IOWR,551) NKIND
+               WRITE(ISHORT,551) NKIND
+  551          FORMAT(/1X,'RUN #',I3,
+     1         '   RESCALE & EVOLVE THE PREVIOUS RUN''S LAST MODEL.')
             END IF
             WRITE(IOWR,452) NMODLS(NKIND),(RESCAL(I,NKIND),I = 1,4)
             WRITE(ISHORT,452) NMODLS(NKIND),

How the outputs show it. In examples/run_from_seed_to_start/m10000gs98z001000_prems, the .short file announces the second card as "RUN # 2 RESCALE THE PREVIOUS RUN'S LAST MODEL." (line 1478↗; line 422 of the file), the message of a card that only rescales. The same .short file and the .track (wrtout.f, line 558↗) show that card's five models, 21 to 25, evolving at its new $Z$ = 0.001: Their ages rise from 5.31806807E-006 to 5.59249319E-006 Gyr, with steps of 52 to 88 years. The first card, also of kind 3, gets its own message, RESCALE & EVOLVE THE INPUT MODEL: "FIRST MODEL". (lines 1473–1475↗).

Checked by running YREC (Nov 9, 2026; a line added at each suspected path counted how often it was taken, the outputs otherwise unchanged): Of YREC's cases with such a card, examples/run_from_seed_to_start/m10000gs98z001000_prems (its run 2) was run, and the wrong message was printed once. With format 551 restored, the track and .store are identical, and only .short changes, by the message. Intent: Format 551 is commented out with no reason given.

2. (changes results, small) Checked by KC (Nov 10, 2026) ISETMIX = 2 and ISETISO = 2 do nothing.

(line 1141↗, line 1149↗, line 1193↗, line 1196↗) The documentation (namelist_control.md, lines 173, 180, and 185) gives 2 as the custom mixture ("CNO enabled if ISETMIX = 2, Light elements if ISETISO=2"); the code takes a custom mixture only with ISETMIX = 1 and AMIX = 'CUS' (ISETISO = 1 and AISO = 'CUS'), and any other value of ISETMIX, ISETISO leaves the starting model's mixture alone without a message. No namelist in the repository sets either to 2.

A possible fix, if the documentation's numbering is the intended one (otherwise, a correction of the documentation):

parmin.f

@@ lines 1138–1144 @@
       ENDIF
       LMIXTURE = .FALSE.
       LISOTOPE = .FALSE.
-      IF(ISETMIX.EQ.1)THEN
+C 11/09/26 KC via Claude
+C code changed to accept ISETMIX = 2 as the custom mixture, as the
+C documentation describes it.
+      IF(ISETMIX.EQ.2) AMIX = 'CUS'
+      IF(ISETMIX.EQ.1 .OR. ISETMIX.EQ.2)THEN
 C IF DEFAULT MIX (GS98) CNOFRACS ARE ALREADY SET.
 C         IF(AMIX.EQ.'GS98')THEN
 C            LMIXTURE = .TRUE.
@@ lines 1190–1196 @@
      *        E12.4,' APPLIED TO STARTING MODEL.')
       ENDIF
 C     CHECK IF ISOTOPE RATIOS NEED TO BE ALTERED
- 602  IF(ISETISO.EQ.1)THEN
+ 602  IF(ISETISO.EQ.2) AISO = 'CUS'
+      IF(ISETISO.EQ.1 .OR. ISETISO.EQ.2)THEN
 C FOR A CUSTOM MIX,DISABLE IF THE SUM OF CNO MASS FRACTIONS EXCEEDS ONE
 C OR IF ANY MASS FRACTION IS NEGATIVE
          IF(AISO.EQ.'CUS')THEN

How the code itself shows it. The code's own comments agree with the code, not with the documentation: MHP's comment of 10/24 gives both ISETISO and ISETMIX as "(1=USED)" and says "AMIX AND AISO ARE STRINGS IDENTIFYING EITHER A PRESET MIXTURE OR A CUSTOM ONE ('CUS')" (lines 502–505↗), and STARIN, which applies the mixture, says "ISETMIX=1 -> CAN ADJUST CNO FRACTIONS ISETISO=1-> CHANGE ISOTOPE RATIOS" (starin.f, line 387↗). The shipped namelists side with the documentation: In 503 of them (in testsuite/, examples/, and sample_cases/), the comments on ISETMIX and ISETISO read "0 = do not adjust 1 = use specified mixture (AMIX below) 2 = custom mixture" (as in testsuite/Test_solarGS98_SF3.nml1, lines 10 and 13), and many add "Custom mixture controls - CNO enabled if ISETMIX = 2, Light elements if ISETISO=2" (line 111 there); none sets AMIX = 'CUS'. So the code and its comments use one numbering, and the documentation and the namelists' comments the other.

Checked by running YREC (Nov 9, 2026; a line added at each suspected path counted how often it was taken, the outputs otherwise unchanged): No case of YREC sets either switch to 2, so testsuite/Test_solarAAG21_dif was run with ISETMIX = ISETISO = 2 and the case's own commented custom-mixture block switched on: The code ignored both, without a message. With the documentation's numbering (the fix), the custom mixture is applied, and the track changes on the pre-main sequence: $\log L$ by up to 0.0025 dex and $\log T_c$ by up to 0.0014 dex (near 0.00018 Gyr), against 0.00023 and 0.000037 dex with a null change (nearly every opacity, all with 3 <= log T <= 9 and -6 <= log rho <= 9.75, times 1 + 1e-10); central $X$ by up to $1.8\times10^{-5}$, within the noise ($3.9\times10^{-5}$). The run keeps its end age, 4.568 Gyr (1,201 models instead of 1,200). The effect is real, though small for this mixture. Intent: The code's comment allows only the default L21 or a custom mixture named by AISO = 'CUS' ("CURRENTLY THERE ARE ONLY 2 VALID OPTIONS"), so the documentation may be what is out of date (question below).

3. (harmless in the runs) ZENV set from every run card.

(line 1486↗) The last line of the card loop, IF(RESCAL(3,NKIND).GE.0.0D0) ZENV=RESCAL(3,NKIND), runs for every card, including evolve cards, where RESCAL is never set (zero), and for rescale cards with the default RSCLZ = 0 ("0 = use current value"). So ZENV leaves PARMIN as the last card's RSCLZ, usually 0. STARIN sets ZENV = ZNEW in the first run (starin.f, line 923↗), and RSCALE sets it whenever it changes $Z$ (rscale.f, lines 137–139↗); in between, the first run's STARIN may call the physics with this ZENV when it moves the fitting point (EQSTAT, GETOPAC, ENVINT, starin.f, lines 641–682↗, starin.f, lines 686–731↗; GETOPAC compares $Z$ with ZENV, getopac.f, line 140↗).

A possible fix, a choice for the developers: drop the line, since RSCALE sets ZENV when it rescales $Z$.

Checked by running YREC (Nov 9, 2026; a line added at each suspected path counted how often it was taken, the outputs otherwise unchanged): The line ran for every run card of all 33 runs, and in 32 of them it set ZENV = 0 from an evolve card. With the line dropped, testsuite/Test_m0p3feh+0p0GS98_scv_ZAMS and the solar calibration examples/run_standard_solar_model/Test_solarcal_noGS_norot give identical tracks and .store files; only .short changes, within 2%. So STARIN's use of this ZENV, if any, changes nothing in these runs, and the item stays fragile code. No comment explains the line.

4. (never occurred) No / in FTRACK leaves LAST_SLASH_IDX unset.

(lines 815–828↗) The loop that finds the last / sets LAST_SLASH_IDX only when there is one. A track file named without a folder leaves it at whatever the saved local holds (in practice 0 at the first call), and mkdir -p "" then fails with a message from the shell; another value would cut FTRACK at a wrong place.

A possible fix:

parmin.f

@@ lines 815–820 @@
 C Create output directory as specified in the FTRACK value of CONTROL
 C namelist if it doesn't already exist.
       ! find index of last '/' char. Use that to snip out the directory name.
+C 11/09/26 KC via Claude
+C code changed to create no folder when FTRACK names none.
+      LAST_SLASH_IDX = 0
       DO I = LEN_TRIM(FTRACK), 1, -1
           IF (FTRACK(I:I) .EQ. '/') THEN
               LAST_SLASH_IDX = I
@@ lines 822–831 @@
           ENDIF
       END DO
 1250  CONTINUE
-      CMD = 'mkdir -p "' // FTRACK(1:LAST_SLASH_IDX) // '"'
-      print *,"OUTPUT placed in :  ",FTRACK(1:LAST_SLASH_IDX)
-      print *, ''
-      CALL system(CMD)
+      IF(LAST_SLASH_IDX.GT.0) THEN
+         CMD = 'mkdir -p "' // FTRACK(1:LAST_SLASH_IDX) // '"'
+         print *,"OUTPUT placed in :  ",FTRACK(1:LAST_SLASH_IDX)
+         print *, ''
+         CALL system(CMD)
+      ENDIF


 C JVS 02/11 Acoustic depth/ Asteroseismic glitch output. Puts output

Checked by running YREC (Nov 9, 2026; a line added at each suspected path counted how often it was taken, the outputs otherwise unchanged): The search ran once in each of the 33 runs, and every FTRACK named a folder (its last / at character 115 to 173), so LAST_SLASH_IDX was always set; the fix was not run. Four namelists of YREC name none: sample_cases/Gyrochrones/rotevolin_m0p2per1p4, rotevolin_m0p25per2p2, rotevolin_m1p25per8p0, and rotevolin_m1p35per8p0 (FTRACK = "rotevolin_m0p2per1p4.track", line 128 of the first); none of them was run, so what they do (whether mkdir -p "" fails) is not checked. No comment explains the missing initialization.

5. (withdrawn) EXPAND_VALUE leaves the end of TEMP undefined.

(line 1527↗, lines 1533–1538↗) Checked, and not a bug: GETENV is called before the test of whether the variable is set (line 1527↗), so in both branches, and when the variable is not set, gfortran's GETENV fills TEMP with blanks (checked with a small program built with the gfortran that builds YREC). In the default branch, TEMP is therefore already blank beyond the expanded name.

Checked by running YREC (Nov 9, 2026; a line added at each suspected path counted how often it was taken, the outputs otherwise unchanged): not reached, since every run set all three variables. The statement rests on gfortran's behavior; a compiler that leaves the argument unchanged for an unset variable would bring the item back.

6. (wrong output) Checked by KC (Nov 10, 2026) Long paths are cut to 256 characters without a word.

(lines 1527–1538↗; lines 825–828↗; pdist.f, lines 86–111↗) Every file name is CHARACTER*256 (lines 80–90↗), and EXPAND_VALUE, which replaces {YREC_OUTPUT} and the other placeholders by their environment variables, never checks the length: A longer result keeps its first 256 characters. Inputs then fail to open (STATUS='OLD'), but outputs are created under the shortened names (STATUS='UNKNOWN'), and names cut inside their shared stem coincide, so the track, short, store, and last files can land in one file. The mkdir command for the output folder (line 825↗) is cut the same way and its failure is ignored, and PDIST's pulsation names lose their model numbers (or, for a name of exactly 256 characters, their folder). Taken with an output folder longer than 182 characters for the repository's longest names; the default (output) is far below it.

A possible fix (a stop; longer names would need every declaration changed, and the mkdir command and PDIST's names, up to 11 and 8 characters longer than the name, would need CMD and COUT lengthened too):

parmin.f

@@ lines 1503–1509 @@
       CHARACTER*256 PHOLDERS(NUM_ENVVARS)
       CHARACTER*256 DEFAULTS(NUM_ENVVARS)
       CHARACTER*256 PHOLDER
-      INTEGER STLEN1, STLEN2, PH_LEN, DEF_LEN
+      INTEGER STLEN1, STLEN2, PH_LEN, DEF_LEN, NEWLEN
 C Each placeholder (env var name enclosed in curly braces) to be
 C supported for expansion in namelists, along with default value in the
 C case where the var is not defined in the execution environment.
@@ lines 1527–1532 @@
             CALL getenv(PHOLDER(2:PH_LEN-1), TEMP)
             STLEN1 = LEN_TRIM(TEMP)
             STLEN2 = LEN_TRIM(NML_VAR)
+C 11/10/26 KC via Claude
+C code changed to stop when the expanded path does not fit in the 256
+C characters of the name (or the variable itself fills them); it was
+C cut short silently, and the run wrote to the shortened names.
+            NEWLEN = DEF_LEN+STLEN2-PH_LEN
+            IF (STLEN1 .NE. 0) NEWLEN = STLEN1+STLEN2-PH_LEN
+            IF (NEWLEN .GT. LEN(NML_VAR) .OR.
+     *          STLEN1 .EQ. LEN(TEMP)) THEN
+               WRITE(*,*) 'EXPAND_VALUE: the path ', NML_VAR(1:STLEN2),
+     *              ' expands to more than ', LEN(NML_VAR),
+     *              ' characters'
+               STOP 3
+            END IF
             IF (STLEN1 .NE. 0) THEN     ! If env var set
                 TEMP(STLEN1+1:STLEN1+STLEN2-PH_LEN) = NML_VAR(PH_LEN+1:STLEN2)
                 !print *,"Override: ",TEMP(1:LEN_TRIM(TEMP))

How the code itself shows it. TEMP(STLEN1+1:STLEN1+STLEN2-PH_LEN) = NML_VAR(PH_LEN+1:STLEN2) (line 1531↗) writes beyond TEMP's 256 characters whenever the sum exceeds 256, and NML_VAR = TEMP (line 1538↗) then cuts; CALL getenv(PHOLDER(2:PH_LEN-1), TEMP) (line 1527↗) already cuts a variable longer than 256. PDIST guards against nothing either: IOCCOL = INDEX(FPMOD,' ')-1 (pdist.f, line 86↗) is $-1$ for a name that fills all 256 characters.

Checked by reading the code (Nov 10, 2026): EXPAND_VALUE expands 23 names (lines 791–813↗), and nothing in it or in its callers compares a name's length with 256. Built alone, with the case's flags and with -fcheck=all, the routine turned {YREC_OUTPUT}/sample_cases/UpperMS/Sample_m9p0feh+0p0GS98_none_TAHB.track with a 201-character YREC_OUTPUT into a name of 256 characters ending in ..._TAHB., with no message and no error in either build, and a separate test showed that gfortran lets two units open one file without an error. The longest output name after its placeholder in the repository's namelists is 74 characters (FSCOMP of testsuite/Test_3Msolar_GS_rot_gray_OPALSCV_SF2_GS98_OP_CF10_RGp2diffrotI3cz2.nml1), so an output folder of more than 182 characters cuts some case's names. In the round of Nov 9, 2026, two fix runs lost their files this way, when the output path exceeded 256 characters. The fix, built into the routine alone, stops (STOP 3, with the message) for the 201-character folder and gives the full 241-character name for a 181-character one. Intent: none found (no comment on the length).

Definitions

subroutinePARMIN

lines 16–1489 · GitHub

Uses modules ACDPTH_MOD, CCOUT_MOD, COMP_MOD, CONSTANTS_MOD, INTPAR_MOD, IOUNITS_MOD, LABEL_MOD, NEWOPAC_MOD, NMLBYPASS_MOD, NMLPAR_MOD, RUNLOOP_MOD

Called by (file level)

External calls (2)

GETARG ×2, SYSTEM

Flowchart of its calls, in order (24)
REMAP775→EXPAND_VALUE791→EXPAND_VALUE792→EXPAND_VALUE793→EXPAND_VALUE794→EXPAND_VALUE795→EXPAND_VALUE796→EXPAND_VALUE797→EXPAND_VALUE798→EXPAND_VALUE799→EXPAND_VALUE800→EXPAND_VALUE801→EXPAND_VALUE802→EXPAND_VALUE803→EXPAND_VALUE804→EXPAND_VALUE805→EXPAND_VALUE806→EXPAND_VALUE807→EXPAND_VALUE808→EXPAND_VALUE809→EXPAND_VALUE810→EXPAND_VALUE811→EXPAND_VALUE812→EXPAND_VALUE813

subroutineEXPAND_VALUE

lines 1494–1541 · GitHub

Replaces a placeholder at the start of a file name, {YREC_INPUT}, {YREC_START} or {YREC_OUTPUT}, by the value of the environment variable of that name, or, if it is not set, by a default: ../../input, ../../startmodels or output. A placeholder elsewhere in the name is left as it is (bug 5, withdrawn, for the default branch).

Called by PARMIN ×23

External calls (1)

GETENV

Change log

The changes since rp_1

Changed since rp_1. The namelists read into modules (NMLPAR_MOD, NMLBYPASS_MOD), where most defaults now are; the removed options are gone (the MHD and OPAL 1995/2001 equations of state, the OPAL92, Kurucz 1990, LAOL89 and Alexander 1995 opacities, Monte Carlo, isochrones), and with them 15 file-name arguments; YRECVER and GITHASH are new arguments, printed and written as a version line; file names may start with a placeholder (EXPAND_VALUE, new), and the output folder is created; the sound-speed file is opened here; the hand-written summary tables are replaced by namelist echoes after ISHORT is opened, with a new DERIVED namelist; TENV is no longer set (ENVINT computes it); the default CMIXL is removed and CMIXLA is checked; PMMM = 0.22 and C_2 = 0.0506 are now double-precision constants; LPMM, PMMWMAX, ZOPAL952 and ZALEX2 are no longer namelist entries. Settled: the namelist dump before OPEN(ISHORT), the printing slip in TRIDT, TRIDL and the stale "standard" values (old questions 1, 3, 4), and part of old question 6 (the unused entries; LSOLWIND now reaches MDOT).

Changed since 09c3b9b. The CONTROL namelist gains L_ZERODT_PREMS (in NMLPAR_MOD, used in main.f), so it is read and echoed to the .short file with the others.

Changed since a8a5950. Two new checks stop the run: LROT with LDIFAD (exit status 1), and a non-positive TRIDT or TRIDL (exit status 3; as the code's comment says, a zero size made SURFBC's reflection walk loop for ever). The stops on bad input (semiconvection with overshoot, a non-positive CMIXLA) now exit with status 3, the code for input errors (KNOWN_ISSUES.md, issue 38).