<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;"><div>Hi,</div><div>now I pretty much see where you are coming from.</div><div>What we have here is a Linux-specific implementation issue (which of course is not going to get fixed anytime soon :-)</div><div><br></div><div>What troubles me a little with the overall idea of this patch is that we are putting a workaround for a lower-level issue in higher-level code.</div><div>If we go this route, the same thing might have to be done for ValueObject::ReadPointedString() which does the same kind of “unbounded read”</div><div>I would prefer, if we do this, that at least we put the code around #if defined (__linux__), as in:</div><div><div> if (error.Fail() || data_read == 0)</div><div> {</div></div><div><div>+ bool found = false;</div></div><div>#if defined (__linux__)</div><div><div>+ if (data_read && error.GetType() == eErrorTypePOSIX) {</div><div>+ // Determine if a null terminator was found in spite of an out-of-bounds read</div><div>+ if (error.GetError() == EIO || error.GetError() == EFAULT) {</div><div>+ char* buffer = reinterpret_cast<char *>(buffer_sp->GetBytes());</div><div>+ char terminator[4] = {'\0', '\0', '\0', '\0'};</div><div>+ assert(sizeof(terminator) >= type_size && "Atempting to read a wchar with more than 4 bytes per character!");</div><div>+</div><div>+ for (size_t i = 0; !found && (i + type_size <= data_read); i += type_size)</div><div>+ if (::strncmp(&buffer[i], terminator, type_size) == 0)</div><div>+ found = true; // null terminator</div><div>+ }</div><div>+ }</div><div><div>#endif</div><div></div></div><div>+ if (!found) {</div></div><div>+ stream.Printf("unable to read data");</div><div><div>+ return true;</div></div><div><div> }</div></div><div>}</div><div><br></div><div>This will maintain a clean behavior for non-Linux systems (and other systems with a similarly broken API can opt-in this behavior by adding themselves to the ifdef)</div><div><br></div><div>But I am inclined to believe the best option would be to adopt your ProcessMonitor fix that reads smaller and smaller chunks and then returns the number of bytes read and no error - that would make all unbounded-read-based calls just work without having to add Linux-specific patches around (main issue being that now we all have to remember that anything that reads a string-like thing has to add a similar workaround or cause Linux unhappiness)</div><div><br></div><div>If that is not going to be viable, or will take some time, I don’t want to keep you blocked, so I would suggest:</div><div>- commit this patch with #ifdef markers to keep it Linux-only</div><div>- file a bugzilla issue to make process reads as functional as possible on Linux</div><div><br></div><div>Thoughts?</div><div><br><div>
<div style="color: rgb(0, 0, 0); letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px; word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;"><div style="font-size: medium; orphans: 2; widows: 2; border-collapse: separate; border-spacing: 0px;"><span style="font-size: 12px; orphans: auto; widows: auto;">Enrico Granata</span><br style="font-size: 12px; orphans: auto; widows: auto;"><span style="font-size: 12px; orphans: auto; widows: auto;">✉ egranata@</span><font color="#ff2600" style="font-size: 12px; orphans: auto; widows: auto;"></font><span style="font-size: 12px; orphans: auto; widows: auto;">.com</span><br style="font-size: 12px; orphans: auto; widows: auto;"><span style="font-size: 12px; orphans: auto; widows: auto;">✆ 27683</span></div></div>
</div>
<br><div><div>On Mar 28, 2013, at 8:39 AM, "Thirumurthi, Ashok" <<a href="mailto:ashok.thirumurthi@intel.com">ashok.thirumurthi@intel.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div lang="EN-US" link="blue" vlink="purple" style="letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;"><div class="WordSection1" style="page: WordSection1;"><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Thanks for the feedback, Enrico,<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">I’m relying on the fact that when the wstring is larger than the default size (1GB in your example), we won’t get a ptrace error like EIO or EFAULT. So, there will be no search for a null terminator.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">If we can’t rely on the null terminator, it’s hard to know if a different error occurred. The trouble is that there is inconsistency in the PTRACE_PEEK implementation on Linux that is more or less arbitrary (sigh), so the out-of-bounds read can generate EIO or EFAULT. In particular, EIO can mean a number of things:<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> “</span><i><span style="font-size: 8pt; font-family: Arial, sans-serif; background-color: white;">request</span></i><span class="apple-converted-space"><span style="font-size: 8pt; font-family: Arial, sans-serif; background-color: white;"> </span></span><span style="font-size: 8pt; font-family: Arial, sans-serif; background-color: white;">is invalid, or an attempt was made to read from or write to an invalid area in the parent's or child's memory, or there was a word-alignment violation, or an invalid signal was specified during a restart request.</span><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">” –<a href="http://linux.die.net/man/2/ptrace" style="color: purple; text-decoration: underline;">http://linux.die.net/man/2/ptrace</a> <o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">In addition, the Linux ProcessMonitor reads in 8-byte chunks, so ReadUTFBufferAndDumpToStream will probably fail (for the right reasons) in the case where at least part of the null terminator was in that block. I see that as an implementation detail that can be resolved in ProcessMonitor by reading smaller chunks until PTRACE_PEEK succeeds (i.e. a good follow-up patch).<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Do let me know if you feel this is okay to commit as is. Cheers,<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt 0.5in; font-size: 12pt; font-family: 'Times New Roman', serif; text-indent: -0.25in;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"><span>-<span style="font-style: normal; font-variant: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-family: 'Times New Roman';"> <span class="Apple-converted-space"> </span></span></span></span><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Ashok<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div><div style="border-style: solid none none; border-top-color: rgb(181, 196, 223); border-top-width: 1pt; padding: 3pt 0in 0in;"><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif;">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif;"><span class="Apple-converted-space"> </span>Enrico Granata [<a href="mailto:egranata@apple.com">mailto:egranata@apple.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Wednesday, March 27, 2013 6:18 PM<br><b>To:</b><span class="Apple-converted-space"> </span>Thirumurthi, Ashok<br><b>Cc:</b><span class="Apple-converted-space"> </span><a href="mailto:lldb-dev@cs.uiuc.edu">lldb-dev@cs.uiuc.edu</a><br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [lldb-dev] PATCH for REVIEW - Fix [Bug 15038] LLDB does not support printing wide-character variables on Linux<o:p></o:p></span></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Hi Ashok,<o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">thanks for submitting an LLDB patch :)<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">I have looked at the code and while I see where the patch is coming from, I see a potential issue with it.<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">The code in ReadUTFBufferAndDumpToStream, which is what you are editing, is meant to work on a partial string (i.e. one that does not have a NULL terminator at the end). The reason for that is that you might have a wstring that is 1GB long and we would not want to try and read all of it and then display it. What we do is pick a size and only extract that much data. For obvious reasons, your string might be longer than our upper boundary, so you would get a chunk of valid bytes and then no end-of-buffer marker.<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">It looks like your code would fail in that case and produce no summary for a string if an EIO was received before \0.<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Is there any reason why you can’t just check if (error == EIO and data_read > 0) and if so treat this as a “partial string” condition and keep going?<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Would that break/crash anything?<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Best,<o:p></o:p></div></div><div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 7pt;">Enrico Granata<br></span><span style="font-size: 7pt; font-family: 'MS Mincho';">✉</span><span style="font-size: 7pt;"><span class="Apple-converted-space"> </span><a href="mailto:egranata@" style="color: purple; text-decoration: underline;">egranata@.com</a><br></span><span style="font-size: 7pt; font-family: 'MS Mincho';">✆</span><span style="font-size: 7pt;"><span class="Apple-converted-space"> </span>27683</span><span style="font-size: 13.5pt;"><o:p></o:p></span></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Mar 27, 2013, at 2:52 PM, "Thirumurthi, Ashok" <<a href="mailto:ashok.thirumurthi@intel.com" style="color: purple; text-decoration: underline;">ashok.thirumurthi@intel.com</a>> wrote:<o:p></o:p></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><br><br><o:p></o:p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">The root cause for Bug 15038 is a ptrace EIO that can occur because we don't know the size of a (UTF) string and so read a fixed number of characters for strings. The attached fix accepts EIO in the case where data has been read and a null terminator of the correct size and alignment was found.<br><br>- Ashok<br><br>-----Original Message-----<br>From:<span class="apple-converted-space"> </span><a href="mailto:lldb-dev-bounces@cs.uiuc.edu" style="color: purple; text-decoration: underline;">lldb-dev-bounces@cs.uiuc.edu</a><span class="apple-converted-space"> </span>[<a href="mailto:lldb-dev-bounces@cs.uiuc.edu" style="color: purple; text-decoration: underline;">mailto:lldb-dev-bounces@cs.uiuc.edu</a>] On Behalf Of<a href="mailto:bugzilla-daemon@llvm.org" style="color: purple; text-decoration: underline;">bugzilla-daemon@llvm.org</a><br>Sent: Tuesday, January 22, 2013 10:13 AM<br>To:<span class="apple-converted-space"> </span><a href="mailto:lldb-dev@cs.uiuc.edu" style="color: purple; text-decoration: underline;">lldb-dev@cs.uiuc.edu</a><br>Subject: [lldb-dev] [Bug 15038] New: LLDB does not support printing wide-character variables on Linux<br><br><a href="http://llvm.org/bugs/show_bug.cgi?id=15038" style="color: purple; text-decoration: underline;">http://llvm.org/bugs/show_bug.cgi?id=15038</a><br><br> Bug #: 15038<br> Summary: LLDB does not support printing wide-character<br> variables on Linux<br> Product: lldb<br> Version: unspecified<br> Platform: PC<br> OS/Version: Linux<br> Status: NEW<br> Severity: enhancement<br> Priority: P<br> Component: All Bugs<br> AssignedTo:<span class="apple-converted-space"> </span><a href="mailto:lldb-dev@cs.uiuc.edu" style="color: purple; text-decoration: underline;">lldb-dev@cs.uiuc.edu</a><br> ReportedBy:<span class="apple-converted-space"> </span><a href="mailto:daniel.malea@intel.com" style="color: purple; text-decoration: underline;">daniel.malea@intel.com</a><br> Classification: Unclassified<br><br><br>Printing a variable of wchar_t type does not behave as expected and results in garbage being printed on the screen.<br><br>To reproduce, remove the @expectedFailureLinux decorator from TestChar1632T.py and TestCxxWCharT.py and run:<br><br>python dotest.py --executable <path-to-lldb> lang/cpp/char1632_t lang/cpp/wchar_t<br><br>--<br>Configure bugmail:<span class="apple-converted-space"> </span><a href="http://llvm.org/bugs/userprefs.cgi?tab=email" style="color: purple; text-decoration: underline;">http://llvm.org/bugs/userprefs.cgi?tab=email</a><br>------- You are receiving this mail because: ------- You are the assignee for the bug.<br>_______________________________________________<br>lldb-dev mailing list<br><a href="mailto:lldb-dev@cs.uiuc.edu" style="color: purple; text-decoration: underline;">lldb-dev@cs.uiuc.edu</a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev" style="color: purple; text-decoration: underline;">http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev</a><br><read-strings.diff>_______________________________________________<br>lldb-dev mailing list<br><a href="mailto:lldb-dev@cs.uiuc.edu" style="color: purple; text-decoration: underline;">lldb-dev@cs.uiuc.edu</a><br><a href="http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev" style="color: purple; text-decoration: underline;">http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev</a></div></div></div></div></div></div></blockquote></div><br></div></body></html>