追踪R segfault的根本原因 [英] Tracing root cause for R segfault

问题描述

我正在使用Shiny和ggplot2编写一个用R编写的非常复杂的仪表板。

现在写了一切正常，但过了一段时间（数据每天更新）它在呈现特定页面时开始出现段错误。

我可以通过打开12个ggplot图形的特定选项卡来重现问题。 b

失败时的控制台输出如下所示：

  dashboard_1 |地址0x194b1，导致'内存未映射'
 dashboard_1 | 
 dashboard_1 |回溯：
 dashboard_1 | 1：mapply（FUN = f，...，SIMPLIFY = FALSE）
 dashboard_1 | 2：地图（build_grob，绘制$ layer，data）
 dashboard_1 | 3：ggplot_gtable（data）
 dashboard_1 | 4：print.ggplot（result $ value）
 dashboard_1 | 5：打印（结果$值）
 dashboard_1 | 6：eval（expr，envir，enclos）
 dashboard_1 | 7：eval（expr，pf）
 dashboard_1 | 8：withVisible（eval（expr，pf））
 dashboard_1 | 9：evalVis（expr）
 dashboard_1 | 10：capture.output（print（result $ value））
 ... 
 dashboard_1 | 65：tryCatchOne（expr，names，parentenv，handlers [[1L]]）
 dashboard_1 | 66：tryCatchList（expr，classes，parentenv，handlers）
 dashboard_1 | 67：tryCatch（（函数（句柄，二进制，消息）{for（处理器在.wsconns [[as.character（句柄）]] $。messageCallbacks）{result<  -  try（handler（binary，message）） if（inherit（result，try-error））{.wsconns [[as.character（handle）]] $ close（）return（）}}}）（62978544，FALSE，{\method \ ：\ update\ \ data\：{\ 日期范围：shiny.date\：[\ 2014-04-26\，\ 2015年\group_by \：\weeks \}}），< environment>），error = .rcpp_error_recorder）
 dashboard_1 | 68：withCallingHandlers（tryCatch（evalq（（function（handle，binary，message）{for（handler in .wsconns [[as.character（handle）]] $。messageCallbacks）{result<  -  try（handler（binary，message ））if（inherits（result，try-error））{.wsconns [[as.character（handle）]] $ close（）return（）}}}）（62978544，FALSE，{\ method\：\ update\，\ data\：{\ 日期范围：shiny.date\：[\ 2014-04-26\，\ 2015-04-26\]，\group_by\：\weeks \}}），< environment>），error = .rcpp_error_recorder），warning = .rcpp_warning_recorder）
 dashboard_1 |正在中止... 
 dashboard_1 |分段错误
 rmreports_dashboard_1已退出，其代码为139 
  

我想追踪根本原因并介绍更改以防止发生段错误。解决这个问题的最好方法是什么？

解决方案

我正在搜索你的问题能够找到这个线程。 https://github.com/hadley/dplyr/issues/322
其中说要尝试用'devtools'更新'dplyr'。
devtools :: install_github（hadley / dplyr，build_vignettes = FALSE）

I'm working on a quite complicated dashboard written in R using Shiny and ggplot2.

At the moment it was written everything worked fine, but after some time (data is updated daily) it started to segfault while rendering specific pages.

I can reproduce the problem by opening specific tab with 12 ggplot graphs.

Console output at the moment of failure looks like this:

dashboard_1 | address 0x194b1, cause 'memory not mapped'
dashboard_1 | 
dashboard_1 | Traceback:
dashboard_1 |  1: mapply(FUN = f, ..., SIMPLIFY = FALSE)
dashboard_1 |  2: Map(build_grob, plot$layer, data)
dashboard_1 |  3: ggplot_gtable(data)
dashboard_1 |  4: print.ggplot(result$value)
dashboard_1 |  5: print(result$value)
dashboard_1 |  6: eval(expr, envir, enclos)
dashboard_1 |  7: eval(expr, pf)
dashboard_1 |  8: withVisible(eval(expr, pf))
dashboard_1 |  9: evalVis(expr)
dashboard_1 | 10: capture.output(print(result$value))
...
dashboard_1 | 65: tryCatchOne(expr, names, parentenv, handlers[[1L]])
dashboard_1 | 66: tryCatchList(expr, classes, parentenv, handlers)
dashboard_1 | 67: tryCatch(evalq((function (handle, binary, message) {    for    (handler in .wsconns[[as.character(handle)]]$.messageCallbacks) {        result <- try(handler(binary, message))        if (inherits(result, "try-error")) {            .wsconns[[as.character(handle)]]$close()            return()        }    }})("62978544", FALSE, "{\"method\":\"update\",\"data\":{\"daterange:shiny.date\":[\"2014-04-26\",\"2015-04-26\"],\"group_by\":\"weeks\"}}"),     <environment>), error = .rcpp_error_recorder)
dashboard_1 | 68: withCallingHandlers(tryCatch(evalq((function (handle, binary,     message) {    for (handler in .wsconns[[as.character(handle)]]$.messageCallbacks) {        result <- try(handler(binary, message))        if (inherits(result, "try-error")) {            .wsconns[[as.character(handle)]]$close()            return()        }    }})("62978544", FALSE, "{\"method\":\"update\",\"data\":{\"daterange:shiny.date\":[\"2014-04-26\",\"2015-04-26\"],\"group_by\":\"weeks\"}}"),     <environment>), error = .rcpp_error_recorder), warning = .rcpp_warning_recorder)
dashboard_1 | aborting ...
dashboard_1 | Segmentation fault
rmreports_dashboard_1 exited with code 139

I would like to trace the root cause and introduce changes to prevent segfault from happening. What is the best way to approach this problem?

解决方案

I was searching your problem was able to find this thread. https://github.com/hadley/dplyr/issues/322 Which says to try to update 'dplyr' with 'devtools'. "devtools::install_github("hadley/dplyr", build_vignettes = FALSE)"

这篇关于追踪R segfault的根本原因的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！