Fix for GitHub issue #377 #378
Conversation
|
There are at least two bugs still lurking in the new code. I plan to add the test cases below, and fix the behavior. For now, I have converted the pull request to a "draft". When the first part of the line read into the buffer ends with the target name, and the first character of the second part of the line is a '#', then the data from the second line part is ignored instead of added to the target name. Observed behavior: Expected behavior: Then there is the problem that a comment in the first part of the line still results in splitting long lines. Observed behavior: Expected behavior: |
|
I plan to first merge #376, then rebase this PR. I would like to merge this PR in about a week, unless there are objections. If you would like me to wait for a positive review of the PR, please let me know. |
* whitespace in sufficiently short lines is skipped * additional words after the target are ignored in sufficiently short lines * whitespace-only (i.e., blank, but not empty) lines are skipped * DOS-format text files, i.e., lines terminated with CRLF, work * input need not be a correct POSIX text file, i.e., non-empty input does not need to end with LF
This addresses GitHub isse schweikert#377. Before, input lines longer than the static buffer used to read data would be handled as if each buffer part were one line. This could result in splitting input target names into multiple different target names. Since the static buffer was significantly smaller then the maximum length of DNS names (just short of 255 one octet characters), valid long DNS names could not be read as targets via file. This commit fixes the splitting of long lines, and also increases the maximum length of target names to 255 one octet characters. The additional line parsing code is kept similar to the existing code and attempts to keep all intended and/or useful features to minimize the risk of regressions.
This pull request has two commits: