edf2csv 0.5.91 → 0.5.93

package/CHANGELOG.md

@@ -3,6 +3,60 @@
 Notable changes to edf2csv. Versions follow [semantic versioning](https://semver.org); while the
 major version is 0, a minor bump may contain breaking changes.
 
+## 0.5.93
+
+### Fixed: the padding after a TAL was counted as an annotation that could not be read
+
+```
+$ edf2csv space-padded.edf --out out
+warning: 2 annotation entries were unreadable and could not be exported.
+
+$ cat out/annotations.csv
+onset_s,duration_s,description,record_index
+0.5,,Lights off,0
+```
+
+The file holds one annotation and it was exported in full. The two "entries" are the spaces
+filling the rest of each record's annotation slot.
+
+EDF+ pads that slot with `0x00`, which the decoder skips because NUL is also what ends one TAL
+and starts the next. Writers pad with spaces, and a run of spaces after the last TAL is a
+non-empty chunk that does not begin with a sign — so it took the malformed branch. Under
+`--strict` that is a failed run over the whitespace at the end of a slot, on a file that lost
+nothing.
+
+Whitespace-only chunks are padding now — space, tab, CR, LF and NUL. A chunk of anything else
+that does not parse is still counted and still reported, which is the case the warning exists
+for.
+
+## 0.5.92
+
+### Fixed: `--info` redirected into a full filesystem wrote nothing and exited 0
+
+```
+$ edf2csv wide.edf --info > desc.txt
+$ echo $?
+0
+$ wc -c desc.txt
+0 desc.txt
+```
+
+A conversion audits what actually reached stdout — that is what 0.5.82 and the disk-image tests
+are about. `--info` wrote its description with `process.stdout.write` and looked at nothing, so
+a run that produced no description at all reported success. `edf2csv rec.edf --info > desc.txt`
+in a script is exactly how someone captures one.
+
+A description is usually a few hundred bytes, which is why this went unnoticed; a 900-channel
+recording's is 58 KB, and no destination is guaranteed to have that.
+
+It uses the same audit now, so it exits 1 and names the cause. That audit declines anything
+that is not a regular file, so a pipe and a terminal are untouched and `--info | head` still
+exits 0.
+
+The test took two attempts: filling the volume completely means the shell cannot create the
+redirect target and the tool never runs, which is what the first one measured. It leaves twenty
+kilobytes and writes 58 into it.
+
 ## 0.5.91
 
 ### Fixed: `--start` at the recording's exact length was accepted when the length was a product

package/dist/cli.js

@@ -18,7 +18,7 @@ import process from 'node:process';
 import { EdfError } from './edf/errors.js';
 import { EdfFile } from './edf/reader.js';
 import { buildPlan, withoutFileRateWarning } from './convert/plan.js';
-import { ConversionError, USAGE_ERROR_CODES, convert, defaultOutputDir, durationDiagnostics, stdoutRefusal, } from './convert/run.js';
+import { ConversionError, USAGE_ERROR_CODES, auditStdout, convert, defaultOutputDir, durationDiagnostics, stdoutRefusal, } from './convert/run.js';
 import { ChannelSelectionError } from './convert/channels.js';
 // Shared with the library so a bad option is the same error whichever way it arrived.
 import { OptionError } from './convert/options.js';
@@ -709,7 +709,21 @@ async function showInfo(input, shared, asJson, jsonIndent, batch = false, toStdo
                 });
             }
         }
-        process.stdout.write(asJson ? `${infoJson(file, plan, jsonIndent)}\n` : `${formatInfo(file, plan)}\n`);
+        /*
+          Checked, like a conversion's stdout is.
+    
+          This wrote and looked at nothing, so `--info > desc.txt` into a filesystem with no room
+          produced a zero-byte file and exited 0. The same audit the `--stdout` path uses: it
+          declines anything that is not a regular file, so a pipe or a terminal is unaffected, and
+          `--info | head` keeps exiting 0.
+        */
+        const audit = auditStdout();
+        const description = asJson
+            ? `${infoJson(file, plan, jsonIndent)}\n`
+            : `${formatInfo(file, plan)}\n`;
+        process.stdout.write(description);
+        audit?.count(Buffer.byteLength(description));
+        audit?.verify();
         // Under --json the warnings travel inside the document, exactly as they do for a
         // conversion, so stderr stays empty and the whole result is one parseable thing.
         const diagnostics = [...withoutFileRateWarning(file.diagnostics), ...plan.diagnostics];