From e380b408fdfafdb3b3e5874edd626cba9a6cbbe3 Mon Sep 17 00:00:00 2001 From: sawinglogz <3787776-sawinglogz@users.noreply.gitlab.com> Date: Mon, 2 Dec 2019 17:30:28 -0500 Subject: [PATCH] Remove warnings about PRS1 oximetry being untested. Also fine-tune various warnings for weird data, such as truncated chunks and multiple sessions in a waveform file. --- oscar/SleepLib/loader_plugins/prs1_loader.cpp | 51 ++++++++++++++++--- oscar/tests/prs1tests.cpp | 4 +- 2 files changed, 45 insertions(+), 10 deletions(-) diff --git a/oscar/SleepLib/loader_plugins/prs1_loader.cpp b/oscar/SleepLib/loader_plugins/prs1_loader.cpp index f4159b2c..2debd6aa 100644 --- a/oscar/SleepLib/loader_plugins/prs1_loader.cpp +++ b/oscar/SleepLib/loader_plugins/prs1_loader.cpp @@ -955,7 +955,6 @@ void PRS1Loader::ScanFiles(const QStringList & paths, int sessionid_base, Machin // All samples exhibiting this behavior are DreamStations. task->m_wavefiles.append(fi.canonicalFilePath()); } else if (ext == 6) { - qWarning() << fi.canonicalFilePath() << "oximetry is untested"; // TODO: mark as untested/unexpected if (!task->oxifile.isEmpty()) { qDebug() << sid << "already has oximetry file" << relativePath(task->oxifile) << "skipping" << relativePath(fi.canonicalFilePath()); @@ -1025,6 +1024,10 @@ void PRS1Loader::ScanFiles(const QStringList & paths, int sessionid_base, Machin //qDebug() << chunkComparison(chunk, task->summary); } else { // Warn about any non-identical duplicate session IDs. + // + // This seems to happen with F5V1 slice 8, which is the only slice in a session, + // and which doesn't update the session ID, so the following slice 7 session + // (which can be hours later) has the same session ID. Neither affects import. qWarning() << chunkComparison(chunk, task->summary); } delete chunk; @@ -7138,13 +7141,27 @@ QList PRS1Import::CoalesceWaveformChunks(QList QString session_s = fi.fileName().section(".", 0, -2); qint32 sid = session_s.toInt(&numeric, m_sessionid_base); if (!numeric || sid != chunk->sessionid) { - qWarning() << chunk->m_path << chunk->sessionid << "session ID mismatch"; + qWarning() << chunk->m_path << "@" << chunk->m_filepos << "session ID mismatch:" << chunk->sessionid; } if (lastchunk != nullptr) { - // Waveform files shouldn't contain multiple sessions + // A handful of 960P waveform files have been observed to have multiple sessions. + // + // This breaks the current approach of deferring waveform parsing until the (multithreaded) + // import, since each session is in a separate import task and could be in a separate + // thread, or already imported by the time it is discovered that this file contains + // more than one session. + // + // For now, we just dump the chunks that don't belong to the session currently + // being imported in this thread, since this happens so rarely. + // + // TODO: Rework the import process to handle waveform data after compliance/summary/ + // events (since we're no longer inferring session information from it) and add it to the + // newly imported sessions. if (lastchunk->sessionid != chunk->sessionid) { - qWarning() << "lastchunk->sessionid != chunk->sessionid in PRS1Loader::CoalesceWaveformChunks()"; + qWarning() << chunk->m_path << "@" << chunk->m_filepos + << "session ID" << lastchunk->sessionid << "->" << chunk->sessionid + << ", skipping" << allchunks.size() - i << "remaining chunks"; // Free any remaining chunks for (int j=i; j < allchunks.size(); ++j) { chunk = allchunks.at(j); @@ -7190,8 +7207,20 @@ QList PRS1Import::CoalesceWaveformChunks(QList } for (int n=0; n < num; n++) { int interleave = chunk->waveformInfo.at(n).interleave; - if (interleave != 5) { - qDebug() << chunk->m_path << "interleave?" << interleave; + switch (chunk->ext) { + case 5: // flow data, 5 samples per second + if (interleave != 5) { + qDebug() << chunk->m_path << "interleave?" << interleave; + } + break; + case 6: // oximetry, 1 sample per second + if (interleave != 1) { + qDebug() << chunk->m_path << "interleave?" << interleave; + } + break; + default: + qWarning() << chunk->m_path << "unknown waveform?" << chunk->ext; + break; } } @@ -7210,6 +7239,7 @@ bool PRS1Import::ParseOximetry() for (int i=0; i < size; ++i) { PRS1DataChunk * oxi = oximetry.at(i); int num = oxi->waveformInfo.size(); + CHECK_VALUE(num, 2); int size = oxi->m_data.size(); if (size == 0) { @@ -7643,8 +7673,15 @@ PRS1DataChunk* PRS1DataChunk::ParseNext(QFile & f) // Make sure the calculated CRC over the entire chunk (header and data) matches the stored CRC. if (chunk->calcCrc != chunk->storedCrc) { - // corrupt data block.. bleh.. + // Correupt data block, warn about it. qWarning() << chunk->m_path << "@" << chunk->m_filepos << "block CRC calc" << hex << chunk->calcCrc << "!= stored" << hex << chunk->storedCrc; + + // TODO: When this happens, it's usually because the chunk was truncated and another chunk header + // exists within the blockSize bytes. In theory it should be possible to rewing and resync by + // looking for another chunk header with the same fileVersion, htype, family, familyVersion, and + // ext (blockSize and other fields could vary). + // + // But this is quite rare, so for now we bail on the rest of the file. break; } diff --git a/oscar/tests/prs1tests.cpp b/oscar/tests/prs1tests.cpp index 1ccb3412..28752eb6 100644 --- a/oscar/tests/prs1tests.cpp +++ b/oscar/tests/prs1tests.cpp @@ -325,9 +325,7 @@ void parseAndEmitChunkYaml(const QString & path) case 1: ok = chunk->ParseSummary(); break; case 2: ok = chunk->ParseEvents(); break; case 5: break; // skip flow/pressure waveforms - case 6: // skip oximetry data (but log it) - qWarning() << relative << "oximetry is untested"; // never encountered - break; + case 6: break; // skip oximetry data default: qWarning() << relative << "unexpected file type"; break;