Three unrelated issues made the musrview_check tests fail on macOS while
passing on Linux.
1. Unspecified argument evaluation order
TColor::GetColor(rand.Integer(255), rand.Integer(255), rand.Integer(255))
evaluates its (side-effecting) arguments in an unspecified order: clang
goes left-to-right, gcc right-to-left, so the same seed yielded RGB on
one platform and BGR on the other. Visible whenever more runs are
plotted than the startup xml colour list provides, e.g.
test-histo-HAL9500.msr with 16 runs against 10 colours.
The rgb values are now drawn into separate variables first. This keeps
the clang result and changes gcc to match, hence the regenerated
reference PNG for musrview-histo-HAL9500 (99.94% of the differing
pixels were exact R<->B swaps).
Fixed in PMusrCanvas, PFourierCanvas and both mupp PMuppCanvas copies.
2. Fixed size path buffers
char fileName[128] plus strncpy(dst, src, sizeof(dst)) truncates *and*
leaves the buffer unterminated once the path reaches the buffer size.
musrview then failed on a 132 character msr-path in doc/examples/ViewOpts.
musrFT and musrt0 were worse: an unbounded strcpy of the startup file
path into char startup_path_name[128], i.e. a stack buffer overflow.
All path/filename buffers in the drivers are now std::string:
musrview, musrFT, musrt0, musrfit, any2many, addRun, dump_header.
PMusrCanvas::SaveGraphicsAndQuit() takes const Char_t* accordingly.
Along the way in the same files:
- msr2msr_replace() wrote a 256 byte line into char temp[128]
- msr2msr assembled "cp"/"rm" shell commands from paths in a 256 byte
buffer; replaced by std::filesystem
- msr2msr and addRun read lines with getline(buf, N), which silently
abandons the rest of the file on the first over-long line
- addRun: bound the unbounded sscanf "%s" to "%255s"
- dropped scratch buffers that only held a string literal, in favour of
TString::ReplaceAll(const char*, const char*)
3. musrview_check.py left its PNGs behind on failure
The two early error returns skipped the cleanup, and since generated
PNGs were identified by "not in the pre-run snapshot", one leftover file
permanently masked the real output of that test: musrview overwrites the
stale PNG, so it was there, just filtered out. One failure thus poisoned
every later run (30 of the 37 observed failures) after a 15 s poll each.
Cleanup now runs in a finally block on every exit path, and generated
PNGs are detected by mtime instead, which sees rewritten leftovers while
staying safe for a sibling test running concurrently under ctest -j.
Also dropped the MUSRVIEW_PNG_DIR env var and its tmp-dir fallback: it
is read nowhere in src/, so the fallback was dead code that guaranteed
the full 15 s poll before every "no PNGs found" failure.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Since the Red/Green mode handling was added for NeXus HDF5 IDF V2, the
histogram numbers of the routed histograms carry the period offset, i.e.
histoNo = period*10000 + histo. PRunBase::DeadTimeCorrection() was still
addressing the dead time parameter vector with 'histoNo-1', hence for any
period > 0 it read far beyond the end of the vector. The garbage picked up
there scaled the counts, so every invocation of musrfit gave a different
(sometimes nan) result for the affected run blocks.
The dead time parameters are stored contiguously, period-by-period, and
therefore need to be addressed as period*noOfHistosPerPeriod + histo - 1.
Introduce PERIOD_HISTO_OFFSET for the period encoding, keep the number of
histos per period in PRawRunData, and add PRawRunData::GetDeadTimeParam(histoNo)
which does the mapping and guards against out-of-range access.
Cross-checked against the HDF4 IDF V1 twin of the same run, where the two
periods are flattened into histos 1-192: both readers now yield bit-identical
chisq, with and without dead time correction.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Uncaught std::bad_any_cast calls terminated the program when an HDF4
attribute's stored type didn't match the expected type (e.g. vector<int>
instead of int, or a non-string units attribute). Wrap all eight attribute
casts with try/catch so type mismatches produce a warning rather than a crash.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
macOS 26 (Darwin 25.5.0) introduced stricter Clang module map rules in
DarwinFoundation1.modulemap: the '_c_standard_library_obsolete' module
now requires the 'found_incompatible_headers__check_search_paths' feature,
which CLING's runtime interpreter does not set. When fftw3.h (from
/opt/homebrew/include) was parsed by CLING via the inlined PStartupHandler
dictionary payload, it triggered this module map error, causing all
PStartupHandler signal/slot connections to fail.
Two fixes:
- PMusr.h: guard '#include "fftw3.h"' with '#ifndef __CLING__'. No fftw
types appear in PMusr.h class definitions, so CLING does not need it
for reflection. This is the primary runtime fix.
- CMakeLists.txt (src/classes and all src/external libs): replace
'-I${FFTW3_INCLUDE}', '-I${Boost_INCLUDE_DIRS}', '-I${GSL_INCLUDE_DIRS}',
and '-I${ROOT_INCLUDE_DIRS}' with '-isystem' in all root_generate_dictionary
OPTIONS blocks, and change 'include_directories(${FFTW3_INCLUDE})' to
'include_directories(SYSTEM ...)'. Internal project paths (MUSRFIT_INC,
BMW_TOOLS_INC, NONLOCAL_INC, CMAKE_CURRENT_SOURCE_DIR, etc.) keep '-I'.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
IsInt now recognises (possibly signed) integers such as "-5" or "+42",
making it slightly more permissive than TString::IsDigit(). A lone sign,
a double sign, or a sign following a digit are still rejected. strToNum
test expectations updated accordingly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ToDouble() had the same shortcoming as ToInt(): wrapping strtod() with a
nullptr endptr made a failed conversion indistinguishable from a
legitimate 0.0. Add an optional `bool *ok` out-parameter that reports
success. strtod() (with endptr + errno) is kept instead of
std::from_chars so the accepted input set stays identical to
TString::Atof() (leading whitespace skipped, leading '+' honoured,
trailing characters ignored); ok is set false on a non-numeric string or
an ERANGE overflow. The parameter defaults to nullptr, so existing call
sites keep compiling unchanged.
Convert the IsFloat-guarded ToDouble call sites in PMsrHandler to the
single-parse ToDouble(token, &ok) form (replacing the IsFloat() guard +
separate ToDouble() that parsed every token twice). All downstream
>=0 / <=0 / range checks are preserved, and push_back sites only append
on success so no spurious 0.0 is stored on error. Number-vs-keyword
discriminators (pos.error/boundary "none", rrf_phase/fourier-phase parX)
are restructured so the keyword branch is taken when ok is false.
As a side effect this fixes a latent gap in the GLOBAL rrf_freq handler,
where a non-numeric frequency previously slipped through with a stale
value instead of raising an error.
The IsFloat-guarded ToInt fit-range offsets (fgb/lgb) are intentionally
left untouched, as there the guard type differs from the conversion.
All 85 integration tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ToInt() previously wrapped strtol() with a nullptr endptr, so a failed
conversion was indistinguishable from a legitimate 0 (matching the old
TString::Atoi() behaviour). Switch the implementation to std::from_chars
and add an optional `bool *ok` out-parameter that reports success: it is
set to false on a non-numeric string or an out-of-range value, true
otherwise. Leading whitespace is skipped and trailing characters are
ignored, preserving the Atoi-like prefix semantics. The parameter
defaults to nullptr, so existing call sites keep compiling unchanged.
Convert the parse-validation call sites in PMsrHandler to the single
-parse ToInt(token, &ok) form, replacing the IsInt() guard + separate
ToInt() (which parsed every token twice). All downstream >0 / >=0 /
range / enum checks are preserved.
Left untouched the call sites where IsInt() acts as a structural
discriminator rather than a numeric validator (write path, xy-data
index-vs-label, fParamInUse usage scans) and the IsFloat-guarded ToInt
offsets, where switching to ToInt(&ok) would change parsing semantics.
All 85 integration tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reduce the ROOT footprint of the MSR parser by removing the pervasive
TString::Tokenize / TObjArray / TObjString / dynamic_cast pattern (28
tokenize sites, 14 TObjArray, 106 TObjString) used to split lines into
tokens, together with the manual `delete tokens` cleanup.
Add a new dependency-free C++17 utility class PStringUtils (Split, IsInt,
IsFloat, ToInt, ToDouble, IsEqualNoCase, ContainsNoCase, BeginsWithNoCase)
that replicates the relevant TString semantics exactly, so it can be reused
elsewhere in the suite. IsInt/IsFloat tolerate surrounding whitespace to
match TString::IsDigit/IsFloat (needed for tokens split on ',' / ';' only).
The public API and the PMusr.h data structures keep TString unchanged; only
the internal tokenizing logic is rewritten. Net -451 lines in
PMsrHandler.cpp. All 85 integration tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
suter_aandClaude Opus 5